NTNU_talk.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "view-in-github",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/howard-haowen/NLP-demos/blob/main/NTNU_talk.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "QEuBPlA2bXZn",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "from traitlets.config.manager import BaseJSONConfigManager\n",
    "from pathlib import Path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "MJFMGDwcba-I",
    "outputId": "58eaa4c3-ba3d-45fc-f41a-973b12a55a84",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'autolaunch': False,\n",
       " 'enable_chalkboard': True,\n",
       " 'scroll': True,\n",
       " 'slideNumber': True,\n",
       " 'controls': True,\n",
       " 'progress': True,\n",
       " 'history': True,\n",
       " 'center': True,\n",
       " 'width': '100%',\n",
       " 'height': '100%',\n",
       " 'theme': 'beige',\n",
       " 'transition': 'concave',\n",
       " 'start_slideshow_at': 'selected'}"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path = Path.home() / \".jupyter\" / \"nbconfig\"\n",
    "cm = BaseJSONConfigManager(config_dir=str(path))\n",
    "cm.update(\n",
    "    \"rise\",\n",
    "    {\n",
    "        \"autolaunch\": False,\n",
    "        \"enable_chalkboard\": True,\n",
    "        \"scroll\": True,\n",
    "        \"slideNumber\": True,\n",
    "        \"controls\": True,\n",
    "        \"progress\": True,\n",
    "        \"history\": True,\n",
    "        \"center\": True,\n",
    "        \"width\": \"100%\",\n",
    "        \"height\": \"100%\",\n",
    "        \"theme\": \"beige\",\n",
    "        \"transition\": \"concave\",\n",
    "        \"start_slideshow_at\": \"selected\"\n",
    "     }\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "pnDIuXs0X3tj",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Outline\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dieJ3Nq4eiSb"
   },
   "source": [
    "- My background\n",
    "- NLP applications in business\n",
    "- Doing linguistic analysis with programs \n",
    "- Machine learning 101\n",
    "- Text representation\n",
    "- Transfer learning\n",
    "- Hands-on walkthrough (HOW)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0s3kuqUKPZta",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# My background"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Ws7aQ1lxX9lq"
   },
   "source": [
    "## Education"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lm5dm2XyaZ9c"
   },
   "source": [
    "![](https://github.com/howard-haowen/NLP-demos/raw/main/img/education_blue.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "x7P84kOpYVxB"
   },
   "source": [
    "## Experience"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KxN1aUc4kr18"
   },
   "source": [
    "- AI Engineer, Wisers Information Limited\n",
    "\n",
    "- AI Engineer, Hamastar Technology\n",
    "\n",
    "- Boya Postdoc Researcher (博雅博士後), Peking University\n",
    "\n",
    "- English Lecturer, National Taipei University of Technology"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "v31_n4ahYbnA",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# NLP applications in business"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "e84jZ0ex8SNV"
   },
   "source": [
    "![](https://camo.githubusercontent.com/7eb96ba311f989fbac5257f412a7f454f695015e7ce52d29ac24e10c8f34ef9e/68747470733a2f2f7777772e63796269616e742e636f6d2f77702d636f6e74656e742f75706c6f6164732f323032302f30312f434b432d4e61747572616c2d4c616e67756167652d50726f63657373696e672e706e67)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dsoHmO9C8C0n"
   },
   "source": [
    "- Document-level sentiment\n",
    "\n",
    "![](https://es.mathworks.com/discovery/sentiment-analysis/_jcr_content/mainParsys/image_695965980_copy.adapt.full.medium.jpg/1623215982364.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "eVpiNzJr84yc"
   },
   "source": [
    "- Category-specific sentiment\n",
    "\n",
    "![](https://www.voxco.com/wp-content/uploads/2021/09/Sentiment-Analysis-helps-improve-Customer-Experience1.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Bjk8H6KgwRBj"
   },
   "source": [
    "- Chatbot\n",
    "\n",
    "![](https://cdn-images-1.medium.com/max/800/1*ILh55uR-pxe9lkotcUXurw.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Ydm28aWc8jbR"
   },
   "source": [
    "- Revenues from the NLP market\n",
    "\n",
    "![](https://github.com/howard-haowen/Chinese-NLP/raw/main/img/nlp_revenues.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hb4xGyyNslMd",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Doing linguistic analysis with programs "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "QlrEqR8WRQAT"
   },
   "source": [
    "![](https://miro.medium.com/max/1200/1*QUNXxPDm5b7CvExfhz9tkg.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6ZZMwASnYnTJ",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Programming languages"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3a3ip6WWSHsA"
   },
   "source": [
    "- Common languages in data science\n",
    "![](https://lh5.googleusercontent.com/rUuG2OeUINVmeaV_OdWJXufjZsSL7EDtPA_TjYC8Mytu1OCmFAlY-fzC4wjzfFyxihcqQMkwJnvnAitqazQeIpr9UvKsBqoLLYDzjSrMSqKToWejWvZ86aTvEGZoHlae1Z5WEdz5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5T4LYdDYXX8D"
   },
   "source": [
    "- Languages across different roles\n",
    "![](https://github.com/howard-haowen/NLP-demos/raw/main/img/languages-in-diff-ds-roles.JPG)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "H1rgxgXYOAeZ"
   },
   "source": [
    "- If you've taken linguistics, you can learn any language, right?\n",
    "- If linguists can decipher an alien language, they can also crack a programming language, right?\n",
    "\n",
    "![](https://i.insider.com/5825e140dd089568398b474c?width=700)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U2q0HQipNwyy",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Developer tools"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "F-40ItZ4ZcCc"
   },
   "source": [
    "### Basic"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Us8FwcHEwJuX"
   },
   "source": [
    "- NLTK, a toolkit for starters\n",
    "![](https://files.realpython.com/media/NLP-for-Beginners-Pythons-Natural-Language-Toolkit-NLTK_Watermarked.16a787c1e9c6.jpg)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zp7TP0yxaI6a"
   },
   "source": [
    "- gensim, a tool that simplifies the process of converting texts to numbers\n",
    "![](https://1.bp.blogspot.com/-zmBmmypQK-I/XVWGnsdXeeI/AAAAAAAABzI/BNAoSQ0sGo0WWih1KvvhXz-uNZYnQ_NvQCLcBGAs/s1600/image001.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ra-ppkiUZwen"
   },
   "source": [
    "- spaCy, an industrial-strength solution for end-to-end pipelines\n",
    "![](https://files.realpython.com/media/How-to-Use-SpaCy-for-Natural-Language-Processing-in-Python_Watermarked_1.b363fc084a80.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "fCmfMZarbOJK"
   },
   "source": [
    "### Advanced"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_GaxajzGlass"
   },
   "source": [
    "- Deep learning frameworks\n",
    "    - `PyTorch`, made by Facebook/Meta\n",
    "    - `TensorFlow`, made by Google\n",
    "    - `Transformers`, made by Hugging Face\n",
    " \n",
    "\n",
    "![](https://venturebeat.com/wp-content/uploads/2019/09/hugging-face.png?w=1200&strip=all)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U8ud-0_KxBWz",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Common NLP tasks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "cvh_Ck57u0jk"
   },
   "source": [
    "### Install tools"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "XoZvcqGNK6mY"
   },
   "source": [
    "- spaCy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "6KLrYo3k8bH-"
   },
   "outputs": [],
   "source": [
    "\n",
    "!pip install -U pip setuptools wheel\n",
    "!pip install -U spacy\n",
    "!python -m spacy download zh_core_web_md"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "wIHSkCqqLwWH"
   },
   "outputs": [],
   "source": [
    "import spacy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3f3xiT7pvflW"
   },
   "source": [
    "- jieba_tw, which is jieba optimized for Taiwan Mandarin"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ZaY5RPFpT8Fq",
    "outputId": "cfcb2037-3800-499a-a654-cb45970a9171"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fatal: destination path 'jieba_tw' already exists and is not an empty directory.\n",
      "/content/jieba_tw\n"
     ]
    }
   ],
   "source": [
    "!git clone -l -s https://github.com/L706077/jieba-zh_TW.git jieba_tw\n",
    "%cd jieba_tw"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "ACh_KTpsUmhC"
   },
   "outputs": [],
   "source": [
    "import jieba"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "5-kTX85mwM20",
    "outputId": "81c3792a-f8a4-42a2-bea5-aa5ca82c80d1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/content\n"
     ]
    }
   ],
   "source": [
    "%cd ../"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Y9YaJKatJ12e"
   },
   "source": [
    "### Tokenization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "aLxNcrXGO7SP"
   },
   "source": [
    "![](https://spacy.io/pipeline-fde48da9b43661abcdf62ab70a546d71.svg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ztHzKCBAfyru",
    "outputId": "2c01ac3e-c915-4a66-e725-333e15a1006d"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['tok2vec', 'tagger', 'parser', 'attribute_ruler', 'ner']"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nlp = spacy.load(\"zh_core_web_md\")\n",
    "nlp.pipe_names"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "duMXgb9K2s4C"
   },
   "source": [
    "- Sample text: `宜家家居新店店店長的名字好長喔！` \n",
    "- This is what inspired me to come up with this example.\n",
    "![](https://www.mirrormedia.com.tw/assets/images/20220123131828-73d4a66ef2493f7ea92c8a28f2ee4f71-tablet.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "G-Z040z2J9-R",
    "outputId": "116400bc-6b5b-4994-8cef-f66ec284a320"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'宜家家 | 居 | 新店 | 店 | 店長 | 的 | 名字 | 好 | 長喔 | ！'"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text = \"宜家家居新店店店長的名字好長喔！\"\n",
    "doc = nlp(text)\n",
    "tokens = [tok.text for tok in doc]\n",
    "\" | \".join(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "JU7p1GVvxSFB",
    "outputId": "15454028-5d34-41aa-cdbc-09df3a8381ea"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'宜家家 | 居 | 新店 | 店 | 店长 | 的 | 名字 | 好长喔 | ！'"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text = \"宜家家居新店店店长的名字好长喔！\"\n",
    "doc = nlp(text)\n",
    "tokens = [tok.text for tok in doc]\n",
    "\" | \".join(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 104
    },
    "id": "l6p3DFIp1rb1",
    "outputId": "39cea788-0f1b-4b6d-fe26-9e4d22bf1afa"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Building prefix dict from the default dictionary ...\n",
      "Loading model from cache /tmp/jieba.cache\n",
      "Loading model cost 0.629 seconds.\n",
      "Prefix dict has been built succesfully.\n"
     ]
    },
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'宜家 | 家居 | 新店店 | 店長 | 的 | 名字 | 好長 | 喔 | ！'"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text = \"宜家家居新店店店長的名字好長喔！\"\n",
    "tokens = jieba.cut(text) \n",
    "\" | \".join(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "collapsed": true,
    "id": "UVAhkqRyVGSr",
    "outputId": "64d17885-5028-4266-f999-e27a5f0a90d6"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'宜家 | 家居 | 新店店 | 店 | 长 | 的 | 名字 | 好 | 长 | 喔 | ！'"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text = \"宜家家居新店店店长的名字好长喔！\"\n",
    "tokens = jieba.cut(text) \n",
    "\" | \".join(tokens)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "j20DhgALCYfY"
   },
   "source": [
    "- The default tokenizer of spaCy is `jieba`, and we'll replace it with `jieba_tw` to get better results on Taiwan Mandarin."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "id": "uifFCEwlw1i6"
   },
   "outputs": [],
   "source": [
    "from spacy.tokens import Doc\n",
    "\n",
    "class TwTokenizer:\n",
    "    def __init__(self, vocab):\n",
    "        self.vocab = vocab\n",
    "\n",
    "    def __call__(self, text):\n",
    "        words =  list(jieba.cut(text))\n",
    "        spaces = [False] * len(words)        \n",
    "        return Doc(self.vocab, words=words, spaces=spaces)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "id": "wRuzILx8yBQw"
   },
   "outputs": [],
   "source": [
    "nlp.tokenizer = TwTokenizer(nlp.vocab)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "N3Hd555gCzpY"
   },
   "source": [
    "- You can replace the sentence stored in `text` with anything off the top of your head."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 35
    },
    "id": "rfI3mLOxxivp",
    "outputId": "d0503557-f058-4f86-e011-c9bec58c28e7"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'宜家 | 家居 | 新店店 | 店長 | 的 | 名字 | 好長 | 喔 | ！'"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text = \"宜家家居新店店店長的名字好長喔！\"\n",
    "doc = nlp(text)\n",
    "tokens = [tok.text for tok in doc]\n",
    "\" | \".join(tokens)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9xeWRF9OzQly"
   },
   "source": [
    "### Part-of-speech (POS) tagger"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wbm4luUV8N9m"
   },
   "source": [
    "![](https://spacy.io/pipeline-fde48da9b43661abcdf62ab70a546d71.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "HxU8Qu_r0RF3"
   },
   "source": [
    "- `Token.pos_` for universal POS tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "R961W_t8zavg",
    "outputId": "f4ec31b8-5064-43a2-d96f-109247725f34"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "宜家 >>> NOUN\n",
      "家居 >>> NOUN\n",
      "新店店 >>> NOUN\n",
      "店長 >>> NOUN\n",
      "的 >>> PART\n",
      "名字 >>> NOUN\n",
      "好長 >>> VERB\n",
      "喔 >>> INTJ\n",
      "！ >>> PUNCT\n"
     ]
    }
   ],
   "source": [
    "for tok in doc:\n",
    "    print(f\"{tok.text} >>> {tok.pos_}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "OcPxI2sy1NdE"
   },
   "source": [
    "- `Token.tag_` for language-specific POS tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "HxerljrF0oy9",
    "outputId": "2b7898b2-8062-4a93-cb31-98f5baade755"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "宜家 >>> NN | noun, singular or mass\n",
      "家居 >>> NN | noun, singular or mass\n",
      "新店店 >>> NN | noun, singular or mass\n",
      "店長 >>> NN | noun, singular or mass\n",
      "的 >>> DEG | associative 的\n",
      "名字 >>> NN | noun, singular or mass\n",
      "好長 >>> VV | other verb\n",
      "喔 >>> IJ | interjection\n",
      "！ >>> PU | punctuation\n"
     ]
    }
   ],
   "source": [
    "for tok in doc:\n",
    "    print(f\"{tok.text} >>> {tok.tag_} | {spacy.explain(tok.tag_)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "veULoMBuyvmw"
   },
   "source": [
    "### Dependency parser"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "EKSNAx1k8Pyz"
   },
   "source": [
    "![](https://spacy.io/pipeline-fde48da9b43661abcdf62ab70a546d71.svg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "id": "2UYOBJ576eVx"
   },
   "outputs": [],
   "source": [
    "from spacy import displacy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 373
    },
    "id": "U32pNPApXs4C",
    "outputId": "5476288e-ddf1-49c6-e14a-a0e6e2291612"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<span class=\"tex2jax_ignore\"><svg xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\" xml:lang=\"zh\" id=\"cb2c2e7b426a4d6ba21129603d319c67-0\" class=\"displacy\" width=\"830\" height=\"332.0\" direction=\"ltr\" style=\"max-width: none; height: 332.0px; color: #000000; background: #ffffff; font-family: Arial; direction: ltr\">\n",
       "<text class=\"displacy-token\" fill=\"currentColor\" text-anchor=\"middle\" y=\"242.0\">\n",
       "    <tspan class=\"displacy-word\" fill=\"currentColor\" x=\"50\">我</tspan>\n",
       "    <tspan class=\"displacy-tag\" dy=\"2em\" fill=\"currentColor\" x=\"50\">PRON</tspan>\n",
       "</text>\n",
       "\n",
       "<text class=\"displacy-token\" fill=\"currentColor\" text-anchor=\"middle\" y=\"242.0\">\n",
       "    <tspan class=\"displacy-word\" fill=\"currentColor\" x=\"180\">想要</tspan>\n",
       "    <tspan class=\"displacy-tag\" dy=\"2em\" fill=\"currentColor\" x=\"180\">VERB</tspan>\n",
       "</text>\n",
       "\n",
       "<text class=\"displacy-token\" fill=\"currentColor\" text-anchor=\"middle\" y=\"242.0\">\n",
       "    <tspan class=\"displacy-word\" fill=\"currentColor\" x=\"310\">三</tspan>\n",
       "    <tspan class=\"displacy-tag\" dy=\"2em\" fill=\"currentColor\" x=\"310\">NUM</tspan>\n",
       "</text>\n",
       "\n",
       "<text class=\"displacy-token\" fill=\"currentColor\" text-anchor=\"middle\" y=\"242.0\">\n",
       "    <tspan class=\"displacy-word\" fill=\"currentColor\" x=\"440\">份</tspan>\n",
       "    <tspan class=\"displacy-tag\" dy=\"2em\" fill=\"currentColor\" x=\"440\">NUM</tspan>\n",
       "</text>\n",
       "\n",
       "<text class=\"displacy-token\" fill=\"currentColor\" text-anchor=\"middle\" y=\"242.0\">\n",
       "    <tspan class=\"displacy-word\" fill=\"currentColor\" x=\"570\">2</tspan>\n",
       "    <tspan class=\"displacy-tag\" dy=\"2em\" fill=\"currentColor\" x=\"570\">NUM</tspan>\n",
       "</text>\n",
       "\n",
       "<text class=\"displacy-token\" fill=\"currentColor\" text-anchor=\"middle\" y=\"242.0\">\n",
       "    <tspan class=\"displacy-word\" fill=\"currentColor\" x=\"700\">號餐</tspan>\n",
       "    <tspan class=\"displacy-tag\" dy=\"2em\" fill=\"currentColor\" x=\"700\">NOUN</tspan>\n",
       "</text>\n",
       "\n",
       "<g class=\"displacy-arrow\">\n",
       "    <path class=\"displacy-arc\" id=\"arrow-cb2c2e7b426a4d6ba21129603d319c67-0-0\" stroke-width=\"2px\" d=\"M70,197.0 C70,132.0 170.0,132.0 170.0,197.0\" fill=\"none\" stroke=\"currentColor\"/>\n",
       "    <text dy=\"1.25em\" style=\"font-size: 0.8em; letter-spacing: 1px\">\n",
       "        <textPath xlink:href=\"#arrow-cb2c2e7b426a4d6ba21129603d319c67-0-0\" class=\"displacy-label\" startOffset=\"50%\" side=\"left\" fill=\"currentColor\" text-anchor=\"middle\">nsubj</textPath>\n",
       "    </text>\n",
       "    <path class=\"displacy-arrowhead\" d=\"M70,199.0 L62,187.0 78,187.0\" fill=\"currentColor\"/>\n",
       "</g>\n",
       "\n",
       "<g class=\"displacy-arrow\">\n",
       "    <path class=\"displacy-arc\" id=\"arrow-cb2c2e7b426a4d6ba21129603d319c67-0-1\" stroke-width=\"2px\" d=\"M330,197.0 C330,67.0 695.0,67.0 695.0,197.0\" fill=\"none\" stroke=\"currentColor\"/>\n",
       "    <text dy=\"1.25em\" style=\"font-size: 0.8em; letter-spacing: 1px\">\n",
       "        <textPath xlink:href=\"#arrow-cb2c2e7b426a4d6ba21129603d319c67-0-1\" class=\"displacy-label\" startOffset=\"50%\" side=\"left\" fill=\"currentColor\" text-anchor=\"middle\">nummod</textPath>\n",
       "    </text>\n",
       "    <path class=\"displacy-arrowhead\" d=\"M330,199.0 L322,187.0 338,187.0\" fill=\"currentColor\"/>\n",
       "</g>\n",
       "\n",
       "<g class=\"displacy-arrow\">\n",
       "    <path class=\"displacy-arc\" id=\"arrow-cb2c2e7b426a4d6ba21129603d319c67-0-2\" stroke-width=\"2px\" d=\"M330,197.0 C330,132.0 430.0,132.0 430.0,197.0\" fill=\"none\" stroke=\"currentColor\"/>\n",
       "    <text dy=\"1.25em\" style=\"font-size: 0.8em; letter-spacing: 1px\">\n",
       "        <textPath xlink:href=\"#arrow-cb2c2e7b426a4d6ba21129603d319c67-0-2\" class=\"displacy-label\" startOffset=\"50%\" side=\"left\" fill=\"currentColor\" text-anchor=\"middle\">mark:clf</textPath>\n",
       "    </text>\n",
       "    <path class=\"displacy-arrowhead\" d=\"M430.0,199.0 L438.0,187.0 422.0,187.0\" fill=\"currentColor\"/>\n",
       "</g>\n",
       "\n",
       "<g class=\"displacy-arrow\">\n",
       "    <path class=\"displacy-arc\" id=\"arrow-cb2c2e7b426a4d6ba21129603d319c67-0-3\" stroke-width=\"2px\" d=\"M590,197.0 C590,132.0 690.0,132.0 690.0,197.0\" fill=\"none\" stroke=\"currentColor\"/>\n",
       "    <text dy=\"1.25em\" style=\"font-size: 0.8em; letter-spacing: 1px\">\n",
       "        <textPath xlink:href=\"#arrow-cb2c2e7b426a4d6ba21129603d319c67-0-3\" class=\"displacy-label\" startOffset=\"50%\" side=\"left\" fill=\"currentColor\" text-anchor=\"middle\">dep</textPath>\n",
       "    </text>\n",
       "    <path class=\"displacy-arrowhead\" d=\"M590,199.0 L582,187.0 598,187.0\" fill=\"currentColor\"/>\n",
       "</g>\n",
       "\n",
       "<g class=\"displacy-arrow\">\n",
       "    <path class=\"displacy-arc\" id=\"arrow-cb2c2e7b426a4d6ba21129603d319c67-0-4\" stroke-width=\"2px\" d=\"M200,197.0 C200,2.0 700.0,2.0 700.0,197.0\" fill=\"none\" stroke=\"currentColor\"/>\n",
       "    <text dy=\"1.25em\" style=\"font-size: 0.8em; letter-spacing: 1px\">\n",
       "        <textPath xlink:href=\"#arrow-cb2c2e7b426a4d6ba21129603d319c67-0-4\" class=\"displacy-label\" startOffset=\"50%\" side=\"left\" fill=\"currentColor\" text-anchor=\"middle\">dobj</textPath>\n",
       "    </text>\n",
       "    <path class=\"displacy-arrowhead\" d=\"M700.0,199.0 L708.0,187.0 692.0,187.0\" fill=\"currentColor\"/>\n",
       "</g>\n",
       "</svg></span>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "text = \"我想要三份2號餐\"\n",
    "doc = nlp(text)\n",
    "displacy.render(doc, style='dep',jupyter=True, options={'distance':130})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "AqRVAlwM6PLM"
   },
   "source": [
    "- `Token.dep_` for dependency tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ASaqLe7i57IQ",
    "outputId": "1212c1bc-9ce8-4dc6-b1fc-5105b1a50de8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "我 >>> nsubj\n",
      "想要 >>> ROOT\n",
      "三 >>> nummod\n",
      "份 >>> mark:clf\n",
      "2 >>> dep\n",
      "號餐 >>> dobj\n"
     ]
    }
   ],
   "source": [
    "for tok in doc:\n",
    "    print(f\"{tok.text} >>> {tok.dep_}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "id": "XykKrZ6I8EpV"
   },
   "outputs": [],
   "source": [
    "from spacy.matcher import Matcher"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "id": "p9VwqL5ff2d0"
   },
   "outputs": [],
   "source": [
    "matcher = Matcher(nlp.vocab)\n",
    "order_pattern = [{\"POS\": \"NUM\", \"OP\": \"+\"}, # for one or more NUM\n",
    "                {\"POS\": \"NOUN\"}]\n",
    "matcher.add(\"ORDER\", [order_pattern])\n",
    "matches = matcher(doc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "1TuTSBuFGnhN",
    "outputId": "c80737ee-6ba5-4dba-d216-9445f5495bdf"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The order is >>> 2號餐\n",
      "The order is >>> 份2號餐\n",
      "The order is >>> 三份2號餐\n"
     ]
    }
   ],
   "source": [
    "matches = matcher(doc)\n",
    "for match_id, start, end in matches:\n",
    "    span = doc[start:end]\n",
    "    print(f\"The order is >>> {span.text}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-W4sPfcZYMCq"
   },
   "source": [
    "### Named entity recognition (NER)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YBk4FJx18vld"
   },
   "source": [
    "![](https://spacy.io/pipeline-fde48da9b43661abcdf62ab70a546d71.svg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 262
    },
    "id": "KozpzTVNYLnb",
    "outputId": "6e6e8e99-53a7-408b-9eab-12cc09ca9566"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<span class=\"tex2jax_ignore\"><div class=\"entities\" style=\"line-height: 2.5; direction: ltr\"></br>（\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    中央社\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "）\n",
       "<mark class=\"entity\" style=\"background: #feca74; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    COVID\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">GPE</span>\n",
       "</mark>\n",
       "-\n",
       "<mark class=\"entity\" style=\"background: #e4e7d2; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    19\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">CARDINAL</span>\n",
       "</mark>\n",
       "疫情使遠距、實體相輔相成的模式成為教學常態，\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    台灣師範大學\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "\n",
       "<mark class=\"entity\" style=\"background: #bfe1d9; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    今天\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">DATE</span>\n",
       "</mark>\n",
       "宣布成立網路大學辦公室，並開設\n",
       "<mark class=\"entity\" style=\"background: #e4e7d2; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    3\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">CARDINAL</span>\n",
       "</mark>\n",
       "個數位碩士在職專班，向全世界招生。</br>\n",
       "<mark class=\"entity\" style=\"background: #feca74; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    台師大\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">GPE</span>\n",
       "</mark>\n",
       "\n",
       "<mark class=\"entity\" style=\"background: #bfe1d9; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    今天\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">DATE</span>\n",
       "</mark>\n",
       "舉辦記者會為網路大學辦公室揭牌，校長吳正己表示，在\n",
       "<mark class=\"entity\" style=\"background: #feca74; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    COVID\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">GPE</span>\n",
       "</mark>\n",
       "-19（\n",
       "<mark class=\"entity\" style=\"background: #bfe1d9; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    2019年\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">DATE</span>\n",
       "</mark>\n",
       "冠狀病毒疾病）疫情下，\n",
       "<mark class=\"entity\" style=\"background: #feca74; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    線上\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">GPE</span>\n",
       "</mark>\n",
       "教學大幅增加，\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    教育部\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "也大幅放寬數位學習學制及招生對象限制。</br></div></span>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "text = \"\"\"\n",
    "（中央社）COVID-19疫情使遠距、實體相輔相成的模式成為教學常態，台灣師範大學今天宣布成立網路大學辦公室，並開設3個數位碩士在職專班，向全世界招生。\n",
    "台師大今天舉辦記者會為網路大學辦公室揭牌，校長吳正己表示，在COVID-19（2019年冠狀病毒疾病）疫情下，線上教學大幅增加，教育部也大幅放寬數位學習學制及招生對象限制。\n",
    "\"\"\"\n",
    "doc = nlp(text)\n",
    "displacy.render(doc, style='ent',jupyter=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "navMRFpm7VDR"
   },
   "source": [
    "- `Span.label_` for NER tags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "6S8gdtjq6v4r",
    "outputId": "77757dea-423a-41dc-cefe-123a7e7e87da"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "中央社 >>> ORG\n",
      "COVID >>> GPE\n",
      "19 >>> CARDINAL\n",
      "台灣師範大學 >>> ORG\n",
      "今天 >>> DATE\n",
      "3 >>> CARDINAL\n",
      "台師大 >>> GPE\n",
      "今天 >>> DATE\n",
      "COVID >>> GPE\n",
      "2019年 >>> DATE\n",
      "線上 >>> GPE\n",
      "教育部 >>> ORG\n"
     ]
    }
   ],
   "source": [
    "for span in doc.ents:\n",
    "    print(f\"{span.text} >>> {span.label_}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "id": "P3RzRy0TlBgZ"
   },
   "outputs": [],
   "source": [
    "from spacy.matcher import PhraseMatcher\n",
    "from spacy.tokens import Span"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "id": "eQBaBH77mO3P"
   },
   "outputs": [],
   "source": [
    "phrase_matcher = PhraseMatcher(nlp.vocab)\n",
    "ntnu_terms = ['台灣師大','台師大',]\n",
    "phrase_matcher.add('NTNU', list(nlp.pipe(ntnu_terms)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 192
    },
    "id": "OOHxLuADmiSV",
    "outputId": "570721ca-77cf-42d8-b569-26d559e751f3"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<span class=\"tex2jax_ignore\"><div class=\"entities\" style=\"line-height: 2.5; direction: ltr\"></br>（\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    中央社\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "）COVID-19疫情使遠距、實體相輔相成的模式成為教學常態，\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    台灣師範大學\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "今天宣布成立網路大學辦公室，並開設3個數位碩士在職專班，向全世界招生。</br>\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    台師大\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "今天舉辦記者會為網路大學辦公室揭牌，校長吳正己表示，在COVID-19（2019年冠狀病毒疾病）疫情下，線上教學大幅增加，\n",
       "<mark class=\"entity\" style=\"background: #7aecec; padding: 0.45em 0.6em; margin: 0 0.25em; line-height: 1; border-radius: 0.35em;\">\n",
       "    教育部\n",
       "    <span style=\"font-size: 0.8em; font-weight: bold; line-height: 1; border-radius: 0.35em; vertical-align: middle; margin-left: 0.5rem\">ORG</span>\n",
       "</mark>\n",
       "也大幅放寬數位學習學制及招生對象限制。</br></div></span>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "matches = phrase_matcher(doc)\n",
    "org_spans = [span for span in doc.ents if span.label_==\"ORG\"]\n",
    "for match_id, start, end in matches:\n",
    "    new_ent = Span(doc, start, end, \"ORG\")\n",
    "    org_spans.append(new_ent)\n",
    "doc.ents = org_spans\n",
    "displacy.render(doc, style='ent',jupyter=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gwy2An4fMD4W"
   },
   "source": [
    "- You can interact with models without knowing any code by using this Web APP that I created.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "id": "7DsB-Rp-MuLF"
   },
   "outputs": [],
   "source": [
    "from IPython.display import IFrame"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "st4Yh1ImNKdP",
    "outputId": "82bede96-c64c-4937-fe47-f5e203919c42"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"700\"\n",
       "            height=\"1000\"\n",
       "            src=\"https://share.streamlit.io/howard-haowen/spacy-streamlit/app.py\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.IFrame at 0x7f322d497890>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "IFrame(src='https://share.streamlit.io/howard-haowen/spacy-streamlit/app.py', width=700, height=1000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Enu3I0--81oS",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Machine learning 101"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "H7sZjiYuEHRJ"
   },
   "source": [
    "- Classical programming vs machine learning\n",
    "\n",
    "![](https://kislayverma.com/wp-content/uploads/2021/06/Screenshot-2021-06-17-at-11.31.04-PM.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "T99aQI8NSQ6a"
   },
   "source": [
    "- ML models learn from instances, so it's not surprising that they make mistakes.\n",
    "\n",
    "![](https://img-9gag-fun.9cache.com/photo/aMKpRdV_460s.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gEN4f_EjDRwz"
   },
   "source": [
    "- AI > ML > DL\n",
    "\n",
    "![](https://monkeylearn.com/static/83ef48b5cf3020a618d878835a6d2167/2948a/image12.webp)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "fodupWuK9tAq"
   },
   "source": [
    "- All forms of data, be it images, sounds, or texts, have to be first converted into numbers to be fed into a training algorithm. \n",
    "- The process of transforming raw data to a sequence of numbers (i.e. vectors) is called `feature engineering`.\n",
    "\n",
    "![](https://miro.medium.com/max/807/1*a-DZhoPMZhONLf-xv7ob8Q.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "FwqWNtdH-3dB"
   },
   "source": [
    "- Two stages of a machine learning project: training and prediction\n",
    " * Taking sentiment analysis as an example\n",
    "\n",
    "![](https://monkeylearn.com/static/74491b164c59b243e5f6d601c0365fbf/76384/ml-process.webp)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "w7MHZyDCJyWM",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Text representation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bP6tFLDBOaAn",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Representing images"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_j3urOZK-eDh"
   },
   "source": [
    "![](http://www.paulvangent.com/wp-content/uploads/2018/08/J2FOX3F.jpg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "S8jUIsPXQLyu"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "collapsed": true,
    "id": "kK9Ony0OQVNZ",
    "outputId": "9073a9e0-6b77-47bb-f211-48a3d11aed58"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2022-03-19 02:04:11--  http://www.paulvangent.com/wp-content/uploads/2018/08/J2FOX3F.jpg\n",
      "Resolving www.paulvangent.com (www.paulvangent.com)... 46.30.213.30, 2a02:2350:5:10a:f2:d80c:fe33:f3de\n",
      "Connecting to www.paulvangent.com (www.paulvangent.com)|46.30.213.30|:80... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 50600 (49K) [image/jpeg]\n",
      "Saving to: ‘sample.jpg’\n",
      "\n",
      "sample.jpg          100%[===================>]  49.41K  93.3KB/s    in 0.5s    \n",
      "\n",
      "2022-03-19 02:04:12 (93.3 KB/s) - ‘sample.jpg’ saved [50600/50600]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "!wget -O sample.jpg http://www.paulvangent.com/wp-content/uploads/2018/08/J2FOX3F.jpg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zMhstbmFQiXB",
    "outputId": "1d6adb75-1b16-4fb4-f57a-0efad5b78488"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(412, 640, 3)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = cv2.imread('sample.jpg')\n",
    "img.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "collapsed": true,
    "id": "0mWsdsN0kq_Z",
    "outputId": "0a973d36-f8f7-452c-cc5d-2e9fa0d30e95"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[245 247 248]\n",
      "  [104 106 106]\n",
      "  [ 50  50  50]\n",
      "  ...\n",
      "  [ 26  26  32]\n",
      "  [ 26  26  32]\n",
      "  [ 27  27  33]]\n",
      "\n",
      " [[240 242 243]\n",
      "  [ 99 101 102]\n",
      "  [ 45  45  45]\n",
      "  ...\n",
      "  [ 19  19  25]\n",
      "  [ 19  19  25]\n",
      "  [ 19  19  25]]\n",
      "\n",
      " [[242 246 247]\n",
      "  [102 106 107]\n",
      "  [ 48  50  50]\n",
      "  ...\n",
      "  [ 20  19  28]\n",
      "  [ 21  20  29]\n",
      "  [ 21  20  29]]\n",
      "\n",
      " ...\n",
      "\n",
      " [[234 243 240]\n",
      "  [ 79  87  86]\n",
      "  [ 15  24  27]\n",
      "  ...\n",
      "  [ 84  84  84]\n",
      "  [ 83  83  83]\n",
      "  [ 82  82  82]]\n",
      "\n",
      " [[235 244 241]\n",
      "  [ 79  87  86]\n",
      "  [ 16  25  28]\n",
      "  ...\n",
      "  [ 84  84  84]\n",
      "  [ 83  83  83]\n",
      "  [ 82  82  82]]\n",
      "\n",
      " [[235 244 241]\n",
      "  [ 80  88  87]\n",
      "  [ 16  25  28]\n",
      "  ...\n",
      "  [ 84  84  84]\n",
      "  [ 83  83  83]\n",
      "  [ 82  82  82]]]\n"
     ]
    }
   ],
   "source": [
    "print(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "X0DBgN0uRCo7",
    "outputId": "274f090c-150c-41b8-9b4e-e268d29f03ff"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(412, 640)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img_gray = cv2.imread('sample.jpg', cv2.IMREAD_GRAYSCALE)\n",
    "img_gray.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 267
    },
    "collapsed": true,
    "id": "scshm1inRLYz",
    "outputId": "5680f515-044c-4f99-8df5-8d7d22b4773e"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(img_gray, cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Zr4iJfuDwEJn"
   },
   "source": [
    "- You can change `100` below to any number between 1 and 255."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 267
    },
    "id": "fYZiIqSwSM9J",
    "outputId": "006f0d1e-7245-4045-df4d-5145b617c12c"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "modified_img = img_gray - 100\n",
    "plt.imshow(modified_img, cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ePaa3CuJhERO",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Representing texts"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wMdcDcD3eNCP"
   },
   "source": [
    "- Compared with images, texts are much harder to represent in numbers.\n",
    "- There're three major types of text representation:\n",
    "    * Symoblic representation\n",
    "    * Distribut**ional** representation\n",
    "    * Distribut**ed** representation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0T_DLHMiFCFn"
   },
   "source": [
    "## Download a dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "collapsed": true,
    "id": "aBe3GwwbFIvR",
    "outputId": "98f3e315-df1c-4936-900c-3b84d407aa13"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2022-03-18 13:00:53--  https://github.com/howard-haowen/NLP-demos/raw/main/Dcard_20220304.db\n",
      "Resolving github.com (github.com)... 140.82.114.4\n",
      "Connecting to github.com (github.com)|140.82.114.4|:443... connected.\n",
      "HTTP request sent, awaiting response... 302 Found\n",
      "Location: https://raw.githubusercontent.com/howard-haowen/NLP-demos/main/Dcard_20220304.db [following]\n",
      "--2022-03-18 13:00:53--  https://raw.githubusercontent.com/howard-haowen/NLP-demos/main/Dcard_20220304.db\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 151552 (148K) [application/octet-stream]\n",
      "Saving to: ‘Dcard.db’\n",
      "\n",
      "Dcard.db            100%[===================>] 148.00K  --.-KB/s    in 0.02s   \n",
      "\n",
      "2022-03-18 13:00:53 (6.74 MB/s) - ‘Dcard.db’ saved [151552/151552]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "!wget -O Dcard.db https://github.com/howard-haowen/NLP-demos/raw/main/Dcard_20220304.db"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "id": "P8NuqQ04FNk3"
   },
   "outputs": [],
   "source": [
    "import sqlite3\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 562
    },
    "id": "89X9EbzhFVFs",
    "outputId": "cef0ad96-ea73-4acf-ed40-8ba572ecdcad"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-c410a9ec-6d62-4af4-a02b-958208c12ccb\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>createdAt</th>\n",
       "      <th>title</th>\n",
       "      <th>excerpt</th>\n",
       "      <th>categories</th>\n",
       "      <th>topics</th>\n",
       "      <th>forum_en</th>\n",
       "      <th>forum_zh</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2022-03-04T07:54:19.886Z</td>\n",
       "      <td>專題需要數據🥺🥺幫填～</td>\n",
       "      <td>希望各位能花個20秒幫我填一下</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2022-03-04T07:42:59.512Z</td>\n",
       "      <td>#詢問 找衣服🥲</td>\n",
       "      <td>想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）</td>\n",
       "      <td>詢問</td>\n",
       "      <td>衣服 | 鞋子 | 衣物 | 男生穿搭 | 尋找</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2022-03-04T07:24:25.147Z</td>\n",
       "      <td>#黑特 網購50% FIFTY PERCENT請三思</td>\n",
       "      <td>因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...</td>\n",
       "      <td></td>\n",
       "      <td>黑特 | 網購 | 三思 | 退貨 | 售後服務</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2022-03-04T06:39:13.017Z</td>\n",
       "      <td>尋衣服</td>\n",
       "      <td>來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎</td>\n",
       "      <td></td>\n",
       "      <td>衣服 | 尋找 | 日常穿搭 | 男生穿搭</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2022-03-04T06:28:06.137Z</td>\n",
       "      <td>#詢問 想問</td>\n",
       "      <td>各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...</td>\n",
       "      <td>詢問</td>\n",
       "      <td>穿搭 | 閒聊版 | 閒聊排解 | 假貨</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-c410a9ec-6d62-4af4-a02b-958208c12ccb')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-c410a9ec-6d62-4af4-a02b-958208c12ccb button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-c410a9ec-6d62-4af4-a02b-958208c12ccb');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                  createdAt                       title  \\\n",
       "0  2022-03-04T07:54:19.886Z                 專題需要數據🥺🥺幫填～   \n",
       "1  2022-03-04T07:42:59.512Z                    #詢問 找衣服🥲   \n",
       "2  2022-03-04T07:24:25.147Z  #黑特 網購50% FIFTY PERCENT請三思   \n",
       "3  2022-03-04T06:39:13.017Z                         尋衣服   \n",
       "4  2022-03-04T06:28:06.137Z                      #詢問 想問   \n",
       "\n",
       "                                             excerpt categories  \\\n",
       "0                                    希望各位能花個20秒幫我填一下              \n",
       "1            想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）         詢問   \n",
       "2  因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...              \n",
       "3                         來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎              \n",
       "4  各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...         詢問   \n",
       "\n",
       "                     topics forum_en forum_zh  \n",
       "0                            dressup       穿搭  \n",
       "1  衣服 | 鞋子 | 衣物 | 男生穿搭 | 尋找  dressup       穿搭  \n",
       "2  黑特 | 網購 | 三思 | 退貨 | 售後服務  dressup       穿搭  \n",
       "3     衣服 | 尋找 | 日常穿搭 | 男生穿搭  dressup       穿搭  \n",
       "4      穿搭 | 閒聊版 | 閒聊排解 | 假貨  dressup       穿搭  "
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conn = sqlite3.connect(\"Dcard.db\")  \n",
    "data = pd.read_sql(\"SELECT * FROM Posts;\", conn)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "kGsoikCRFnqw",
    "outputId": "241b5804-4361-492f-d7ac-cca2924cb89a"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "forum_en\n",
       "dressup         30\n",
       "food            30\n",
       "funny           30\n",
       "girl            30\n",
       "horoscopes      30\n",
       "makeup          30\n",
       "money           30\n",
       "netflix         30\n",
       "relationship    30\n",
       "talk            30\n",
       "trending        30\n",
       "youtuber        30\n",
       "dtype: int64"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby(\"forum_en\").size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "_8d4VTn0xTBd",
    "outputId": "4eccc09f-fe98-4bac-c24d-7a46d6ded067"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(360, 7)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 606
    },
    "id": "_CEFl2g3GSiT",
    "outputId": "00b87293-d310-47be-d70a-adaba5a0f5d1"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-161a539b-23e2-40af-8364-1df3ca274c09\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>excerpt</th>\n",
       "      <th>forum_en</th>\n",
       "      <th>forum_zh</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>希望各位能花個20秒幫我填一下</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>355</th>\n",
       "      <td>昨天上了第一支影片，之前有發過沒有線條的動畫影片，新的頻道改成有線條的，感覺大家好像比較喜歡...</td>\n",
       "      <td>youtuber</td>\n",
       "      <td>YouTuber</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>356</th>\n",
       "      <td>今天全台灣大停電，應該過幾天就會有個戴面具的出來說，一定是中共……，我從上個影片就預測了……</td>\n",
       "      <td>youtuber</td>\n",
       "      <td>YouTuber</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>357</th>\n",
       "      <td>想問有沒有人知道阿神和放火是認識還是有結過什麼仇之類的嗎？首先我個人基本沒關注過放火，但是最...</td>\n",
       "      <td>youtuber</td>\n",
       "      <td>YouTuber</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>358</th>\n",
       "      <td>無意引戰，單純分享我的觀察與個人想法～這幾天看了Dcard幾篇關於Rice&amp; Shine的貼...</td>\n",
       "      <td>youtuber</td>\n",
       "      <td>YouTuber</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>359</th>\n",
       "      <td>哈哈哈哈，沒錯我就是親友團來介紹一個我覺得很北七的頻道，現在觀看真的低的可憐，也沒事啦，就多...</td>\n",
       "      <td>youtuber</td>\n",
       "      <td>YouTuber</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>360 rows × 3 columns</p>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-161a539b-23e2-40af-8364-1df3ca274c09')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-161a539b-23e2-40af-8364-1df3ca274c09 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-161a539b-23e2-40af-8364-1df3ca274c09');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                               excerpt  forum_en  forum_zh\n",
       "0                                      希望各位能花個20秒幫我填一下   dressup        穿搭\n",
       "1              想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）   dressup        穿搭\n",
       "2    因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...   dressup        穿搭\n",
       "3                           來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎   dressup        穿搭\n",
       "4    各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...   dressup        穿搭\n",
       "..                                                 ...       ...       ...\n",
       "355  昨天上了第一支影片，之前有發過沒有線條的動畫影片，新的頻道改成有線條的，感覺大家好像比較喜歡...  youtuber  YouTuber\n",
       "356     今天全台灣大停電，應該過幾天就會有個戴面具的出來說，一定是中共……，我從上個影片就預測了……  youtuber  YouTuber\n",
       "357  想問有沒有人知道阿神和放火是認識還是有結過什麼仇之類的嗎？首先我個人基本沒關注過放火，但是最...  youtuber  YouTuber\n",
       "358  無意引戰，單純分享我的觀察與個人想法～這幾天看了Dcard幾篇關於Rice& Shine的貼...  youtuber  YouTuber\n",
       "359  哈哈哈哈，沒錯我就是親友團來介紹一個我覺得很北七的頻道，現在觀看真的低的可憐，也沒事啦，就多...  youtuber  YouTuber\n",
       "\n",
       "[360 rows x 3 columns]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols = [\"excerpt\", \"forum_en\", \"forum_zh\"]\n",
    "df = data[cols]\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2mMH2tvTDSIE"
   },
   "source": [
    "- You can change `62` below to any number between 0 and 359."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 52
    },
    "id": "f8lT2sGwljIQ",
    "outputId": "942f50e2-4198-49fb-d02f-bc2e1441eb3f"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "type": "string"
      },
      "text/plain": [
       "'昨天在屈臣氏買的新品，手刀來分享心得！！！！，🥑價格： $305/320ml～～很大條，成分主打維生素B群、紅石榴跟枸杞，跟其他兩款一樣都是果萃的美白，這款有很少女的莓果香，甜甜的感覺，味道有點像美國'"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "docid = 62\n",
    "sample_text = df.loc[docid, \"excerpt\"]\n",
    "sample_text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "znoBLKfLHGu0"
   },
   "source": [
    "## Preprocess texts"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-0bCdr8PRdYM"
   },
   "source": [
    "![](https://res.cloudinary.com/practicaldev/image/fetch/s--95IEaI7P--/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_880/https://dev-to-uploads.s3.amazonaws.com/uploads/articles/x9qz2q3gyz494rwz3i7q.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "id": "b60YVo8MGhW1"
   },
   "outputs": [],
   "source": [
    "def preprocess_text(text: str) -> list:\n",
    "    doc = nlp.make_doc(text)\n",
    "    res = [tok for tok in doc if not tok.is_punct]\n",
    "    res = [tok for tok in res if not tok.is_stop] # stop words\n",
    "    res = [tok for tok in res if not tok.like_email]\n",
    "    res = [tok for tok in res if not tok.like_url]\n",
    "    res = [tok for tok in res if not tok.like_num]\n",
    "    res = [tok for tok in res if not tok.is_ascii]\n",
    "    res = [tok.text for tok in res if not tok.is_space]\n",
    "    return res"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "wPDj6uSVEtSZ"
   },
   "source": [
    "- Think of stop words as highly frequent words, typically those that non-native speakers tend to make mistakes on.\n",
    "\n",
    "![](https://www.computerhope.com/jargon/s/stop-words.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "fiwrzUsJkGgG",
    "outputId": "b0a3a909-b9b9-4117-8753-a87e4ba6fa43"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Before preprocessing:\n",
      " 昨天在屈臣氏買的新品，手刀來分享心得！！！！，🥑價格： $305/320ml～～很大條，成分主打維生素B群、紅石榴跟枸杞，跟其他兩款一樣都是果萃的美白，這款有很少女的莓果香，甜甜的感覺，味道有點像美國\n",
      "After preprocessing:\n",
      " ['昨天', '屈臣氏', '買', '新品', '手刀', '來', '分享', '心得', '🥑', '價格', '條', '成分', '主打', '維生素', 'B群', '紅', '石榴', '枸杞', '兩', '款', '一樣', '果萃', '美白', '這', '款有', '少女', '莓果', '香', '甜甜', '感覺', '味道', '有點', '美國']\n"
     ]
    }
   ],
   "source": [
    "processed_text = preprocess_text(sample_text)\n",
    "print(f\"Before preprocessing:\\n {sample_text}\")\n",
    "print(f\"After preprocessing:\\n {processed_text}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "gDveHNeoGu0W"
   },
   "outputs": [],
   "source": [
    "df.loc[:,\"toks\"] = df[\"excerpt\"].apply(preprocess_text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 371
    },
    "id": "slgYRqT6jkWD",
    "outputId": "24956a51-724b-499f-c5a1-c081bc25d90d"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-1f13b3c3-5c8f-4fc6-a7b5-94a92b30bc73\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>excerpt</th>\n",
       "      <th>forum_en</th>\n",
       "      <th>forum_zh</th>\n",
       "      <th>toks</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>希望各位能花個20秒幫我填一下</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[希望, 位, 花, 個, 秒, 幫, 填]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[想, 找, 這, 套, 衣服, 🥲, 發現, 該, 用什麼, 關鍵字, 找, 圖是, 草屯...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[因為, 文會, 有點, 長, 先, 說, 結論, 網購, 過, 平台, 退貨, 麻煩, 家...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[來源, 覺得, 呱吉, 這, 襯衫, 好好, 類似, 嗎]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[位, 因為, 這, 個, 證件, 夾, 臺灣, 買不到, 美國, 限量版, 貨, 這, 間...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-1f13b3c3-5c8f-4fc6-a7b5-94a92b30bc73')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-1f13b3c3-5c8f-4fc6-a7b5-94a92b30bc73 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-1f13b3c3-5c8f-4fc6-a7b5-94a92b30bc73');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                             excerpt forum_en forum_zh  \\\n",
       "0                                    希望各位能花個20秒幫我填一下  dressup       穿搭   \n",
       "1            想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）  dressup       穿搭   \n",
       "2  因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...  dressup       穿搭   \n",
       "3                         來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎  dressup       穿搭   \n",
       "4  各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...  dressup       穿搭   \n",
       "\n",
       "                                                toks  \n",
       "0                             [希望, 位, 花, 個, 秒, 幫, 填]  \n",
       "1  [想, 找, 這, 套, 衣服, 🥲, 發現, 該, 用什麼, 關鍵字, 找, 圖是, 草屯...  \n",
       "2  [因為, 文會, 有點, 長, 先, 說, 結論, 網購, 過, 平台, 退貨, 麻煩, 家...  \n",
       "3                     [來源, 覺得, 呱吉, 這, 襯衫, 好好, 類似, 嗎]  \n",
       "4  [位, 因為, 這, 個, 證件, 夾, 臺灣, 買不到, 美國, 限量版, 貨, 這, 間...  "
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "woJHLfkV1FyS",
    "outputId": "537d481f-81a4-47ba-a074-47bdd05f0752"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['dressup', 'relationship', 'makeup', 'food', 'horoscopes', 'talk',\n",
       "       'trending', 'money', 'funny', 'girl', 'netflix', 'youtuber'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"forum_en\"].unique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "VkU0l9KzwwWX"
   },
   "source": [
    "- You can choose any forum name from above and append it to `selected_forums`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "id": "avm5rASl0-x-"
   },
   "outputs": [],
   "source": [
    "selected_forums = [\"trending\", \"netflix\", \"money\", \"makeup\", \"dressup\"]\n",
    "forum_filt = df[\"forum_en\"].isin(selected_forums)\n",
    "sample_df = df[forum_filt]\n",
    "sample_df.reset_index(inplace=True, drop=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_TxQWSqxIKvu"
   },
   "source": [
    "- One common way to get rid of outliers is to filter by text length. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 363
    },
    "id": "aZLRVURw9Ylx",
    "outputId": "8a8bcaa2-d22f-419d-a735-0782b9e90d81"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-9df27619-dbba-4f04-bfe6-151216eefc22\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>index</th>\n",
       "      <th>excerpt</th>\n",
       "      <th>forum_en</th>\n",
       "      <th>forum_zh</th>\n",
       "      <th>toks</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>8</td>\n",
       "      <td></td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>29</td>\n",
       "      <td></td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>69</td>\n",
       "      <td>是在笑屁笑啦🤷‍️</td>\n",
       "      <td>makeup</td>\n",
       "      <td>美妝</td>\n",
       "      <td>[笑, 屁笑, 🤷, ‍, ️]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>58</th>\n",
       "      <td>88</td>\n",
       "      <td>如題，請教各位</td>\n",
       "      <td>makeup</td>\n",
       "      <td>美妝</td>\n",
       "      <td>[如題, 請教, 位]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66</th>\n",
       "      <td>186</td>\n",
       "      <td></td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>78</th>\n",
       "      <td>198</td>\n",
       "      <td>感覺離戰爭好近</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>[感覺, 離, 戰爭, 好近]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>82</th>\n",
       "      <td>202</td>\n",
       "      <td></td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>88</th>\n",
       "      <td>208</td>\n",
       "      <td>金曲不敗</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>[金曲, 敗]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>100</th>\n",
       "      <td>220</td>\n",
       "      <td></td>\n",
       "      <td>money</td>\n",
       "      <td>理財</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>144</th>\n",
       "      <td>324</td>\n",
       "      <td></td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-9df27619-dbba-4f04-bfe6-151216eefc22')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-9df27619-dbba-4f04-bfe6-151216eefc22 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-9df27619-dbba-4f04-bfe6-151216eefc22');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "     index    excerpt  forum_en forum_zh              toks\n",
       "8        8              dressup       穿搭                []\n",
       "29      29              dressup       穿搭                []\n",
       "39      69  是在笑屁笑啦🤷‍️    makeup       美妝  [笑, 屁笑, 🤷, ‍, ️]\n",
       "58      88    如題，請教各位    makeup       美妝       [如題, 請教, 位]\n",
       "66     186             trending       時事                []\n",
       "78     198    感覺離戰爭好近  trending       時事   [感覺, 離, 戰爭, 好近]\n",
       "82     202             trending       時事                []\n",
       "88     208       金曲不敗  trending       時事           [金曲, 敗]\n",
       "100    220                money       理財                []\n",
       "144    324              netflix  Netflix                []"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_df[sample_df[\"excerpt\"].apply(len)<=10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "id": "0mO5oNuS-HOC"
   },
   "outputs": [],
   "source": [
    "sample_df = sample_df[sample_df[\"excerpt\"].apply(len)>10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "id": "-NzolM5A-b3E"
   },
   "outputs": [],
   "source": [
    "sample_df.reset_index(inplace=True, drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "LD9BPF-F-VY-",
    "outputId": "f7d37a0d-cb28-444c-8c89-b1f541bcc76b"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(140, 6)"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3lXNDkIE3ryo",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Symbolic representation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "OkXUDKlaJ7Z_"
   },
   "source": [
    "### TF-IDF (term frequency-inverse document frequency)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "l2u951-CQKxA"
   },
   "source": [
    "![](https://miro.medium.com/max/1000/1*vWWmJlDykVRkjg9c38VbxQ.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dBVz4zFz4Iix"
   },
   "source": [
    "- TF-IDF is based on the Bag-of-Words (BOW) model.\n",
    "![](https://okan.cloud/posts/2021-04-08-text-vectorization-using-python-term-document-matrix/bow.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "id": "4RzA97fLWdOn"
   },
   "outputs": [],
   "source": [
    "from gensim import corpora, models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "JvjqoVDqxWBZ"
   },
   "source": [
    "- A gensim dictionary stores a mapping between words and their integer IDs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Gs7idfgFvm77",
    "outputId": "b3f012ea-f7c2-4f34-8678-41d47e5ed1ab"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'位': 0, '個': 1, '填': 2, '希望': 3, '幫': 4, '秒': 5, '花': 6, '囝仔': 7, '圖是': 8, '套': 9, '想': 10, '截圖': 11, '找': 12, '校園': 13, '演唱會': 14, '用什麼': 15, '發現': 16, '草屯': 17, '衣服': 18, '該': 19, '這': 20, '關鍵字': 21, '\\U0001f972': 22, '以為': 23, '保障': 24, '先': 25, '刁難': 26, '刻意': 27, '原本': 28, '售後': 29, '因為': 30, '家': 31, '實體店': 32, '平台': 33, '念頭': 34, '打消': 35, '文會': 36, '會': 37, '有點': 38, '服務': 39, '正': 40, '比較': 41, '沒想到': 42, '糟': 43, '結論': 44, '網購': 45, '認為': 46, '說': 47, '讓': 48, '退貨': 49, '遇過': 50, '過': 51, '長': 52, '顧客': 53, '麻煩': 54, '來源': 55, '呱吉': 56, '嗎': 57, '好好': 58, '襯衫': 59, '覺得': 60, '類似': 61, '不會': 62, '假': 63, '夾': 64, '美國': 65, '臺灣': 66, '蝦皮': 67, '請': 68, '謝謝': 69, '證件': 70, '貨': 71, '買': 72, '買不到': 73, '買到': 74, '這會': 75, '還是': 76, '間': 77, '限量版': 78, '下單': 79, '什麼': 80, '入手': 81, '取消': 82, '後': 83, '斷貨': 84, '方式': 85, '訂單': 86, '請問': 87, '賣': 88, '雙鞋': 89, '´': 90, '▽': 91, '們': 92, '分享': 93, '到來': 94, '厚重': 95, '哈囉': 96, '單品': 97, '囉': 98, '大使': 99, '小魚': 100, '已經': 101, '放到': 102, '春夏': 103, '春天': 104, '歐膩': 105, '溫柔': 106, '為了': 107, '現在': 108, '穿搭': 109, '萬年': 110, '衣櫃': 111, '裡面': 112, '迎接': 113, '開箱': 114, '靴子': 115, 'ﾉ': 116, '件': 117, '來問': 118, '出來': 119, '實體': 120, '店面': 121, '想問': 122, '改短': 123, '有夠': 124, '為什麼': 125, '燒到': 126, '穿': 127, '網頁': 128, '褲子': 129, '試穿': 130, '請人': 131, '資訊': 132, '跳': 133, '近期': 134, '逛過': 135, '點': 136, '一場': 137, '三天': 138, '偏': 139, '做': 140, '兩': 141, '出遊': 142, '出門': 143, '參考': 144, '墾丁': 145, '夜': 146, '太多': 147, '帶': 148, '旅': 149, '旅程': 150, '東西': 151, '畢竟': 152, '簡單': 153, '色系': 154, '記錄': 155, '走': 156, '踏上': 157, '雖然': 158, '黑白': 159, '\\U0001f90d': 160, '代購': 161, '叫壯': 162, '小弟': 163, '怪怪': 164, '推薦': 165, '有沒有': 166, '服飾': 167, '男生': 168, '美系': 169, '賣場': 170, '韓版': 171, '體型': 172, '吸引': 173, '哪裡': 174, '問': 175, '弱弱': 176, '狐狸': 177, '脆脆': 178, '貓咪': 179, '些賣': 180, '假貨': 181, '包包': 182, '正品': 183, '直播': 184, '精品': 185, '上衣': 186, '下架': 187, '剛剛': 188, '客廳': 189, '本來': 190, '結果': 191, '詢問': 192, '🥺': 193, '女生': 194, '挺': 195, '推出': 196, '款': 197, '短袖': 198, '胸部': 199, '適合': 200, '一樣': 201, '不用': 202, '中': 203, '喜歡': 204, '塊': 205, '太': 206, '小腿': 207, '微肉': 208, '擺盪': 209, '用力': 210, '瘦子': 211, '發達': 212, '肌': 213, '胖': 214, '身材': 215, '身高': 216, '長靴': 217, '體重': 218, '丟': 219, '半夜': 220, '外套': 221, '天': 222, '實現': 223, '幾': 224, '想要': 225, '找到': 226, '整整': 227, '朋友': 228, '沒': 229, '沒有': 230, '消息': 231, '男友': 232, '網路': 233, '購買': 234, '還': 235, '難以': 236, '大四': 237, '好看': 238, '學生': 239, '選手': 240, '偶然': 241, '品牌': 242, '大戰': 243, '好心人': 244, '如題': 245, '支持': 246, '星際': 247, '正版': 248, '爬文': 249, '經': 250, '表弟': 251, '評價': 252, '送': 253, '選擇性': 254, '好了': 255, '實在': 256, '寬褲': 257, '平常': 258, '怎麼': 259, '搭': 260, '條': 261, '照片': 262, '種': 263, '練習': 264, '西裝': 265, '這樣': 266, '配了': 267, '風格': 268, '來': 269, '其實': 270, '前': 271, '十年': 272, '卡包': 273, '外': 274, '扣除': 275, '掉': 276, '次': 277, '版開': 278, '男夾': 279, '當': 280, '短夾': 281, '禮物': 282, '算是': 283, '耐操': 284, '超': 285, '零錢包': 286, '努力': 287, '女兒': 288, '媽媽': 289, '少女': 290, '幾乎': 291, '幾年': 292, '懶': 293, '手機': 294, '打扮': 295, '拍照': 296, '放棄': 297, '時期': 298, '熟女': 299, '生完': 300, '維持': 301, '裡': 302, '規格': 303, '越來越': 304, '邋遢': 305, '˂': 306, '˶': 307, '̗': 308, '̀': 309, 'ᴗ': 310, 'ᵕ': 311, '三月': 312, '不知不覺': 313, '來到': 314, '咿': 315, '嗨': 316, '回暖': 317, '天氣': 318, '嫌棄': 319, '我們': 320, '漸漸': 321, '看下去': 322, '秋冬': 323, '繼續': 324, '話': 325, '𝚌': 326, '𝚐': 327, '𝚔': 328, '𝚖': 329, '🧸': 330, '二手': 331, '五分': 332, '全新': 333, '官網': 334, '直紋': 335, '聯名': 336, '號': 337, '難過': 338, '不錯': 339, '他們': 340, '年初': 341, '無腦': 342, '版型': 343, '產品': 344, '瘋狂': 345, '真的': 346, '褲': 347, '買下': 348, '追蹤': 349, '這不': 350, '韓貨': 351, '黑色': 352, '牌子': 353, '這是': 354, 'ㄉ': 355, '身邊': 356, '類': 357, '˵': 358, '̫': 359, '᷄': 360, '᷅': 361, '⁻': 362, '好人': 363, '片': 364, '介紹': 365, '光感': 366, '全包': 367, '哭出': 368, '唇彩': 369, '唇膏': 370, '唇釉': 371, '支': 372, '沉迷': 373, '空氣': 374, '絲絨': 375, '要來': 376, '變成': 377, '那麼': 378, '錢包': 379, '開玩笑': 380, '中國': 381, '口紅': 382, '堆': 383, '大開眼界': 384, '寶庫': 385, '小紅書': 386, '想到': 387, '搜尋': 388, '淘寶': 389, '生火': 390, '眼影': 391, '笑': 392, '腮紅': 393, '開始': 394, 'B群': 395, '主打': 396, '價格': 397, '味道': 398, '屈臣氏': 399, '心得': 400, '感覺': 401, '成分': 402, '手刀': 403, '新品': 404, '昨天': 405, '果萃': 406, '枸杞': 407, '款有': 408, '甜甜': 409, '石榴': 410, '紅': 411, '維生素': 412, '美白': 413, '莓果': 414, '香': 415, '🥑': 416, '優惠': 417, '克蘭詩': 418, '入坑': 419, '印象': 420, '專屬': 421, '常會': 422, '折扣': 423, '林三益': 424, '玄玄': 425, '發': 426, '瞜': 427, '碼': 428, '蘭蔻': 429, '蠻': 430, '長會': 431, '限定組': 432, '隱藏': 433, '雅勻': 434, '一門': 435, '不厭': 436, '保養品': 437, '刺刺': 438, '大致上': 439, '妖豔': 440, '學問': 441, '小心翼翼': 442, '擦上': 443, '放心': 444, '敏感': 445, '服用': 446, '泛紅': 447, '濃郁': 448, '牌': 449, '百試': 450, '立馬': 451, '耶': 452, '醫美': 453, '香味': 454, '交友': 455, '充滿': 456, '南部': 457, '台積電': 458, '同事': 459, '單身': 460, '壓縮機': 461, '工程師': 462, '揣測': 463, '日本': 464, '根據': 465, '歲': 466, '濕氣': 467, '無塵': 468, '稀少': 469, '統計': 470, '衣裡': 471, '認識': 472, '透過': 473, '進口': 474, '一陣子': 475, '乾性': 476, '保養': 477, '刺激': 478, '嚴重': 479, '忽冷': 480, '忽熱': 481, '摧殘': 482, '狀態': 483, '穩定': 484, '肌膚': 485, '臉頰': 486, '變得': 487, '轉用': 488, '🆘': 489, '付款': 490, '以來': 491, '力點': 492, '噴': 493, '困擾': 494, '地方': 495, '嫩粉': 496, '專業': 497, '對話': 498, '從': 499, '截': 500, '排版': 501, '接觸': 502, '擦': 503, '文': 504, '是業': 505, '桃色': 506, '正文': 507, '烏唇': 508, '發文': 509, '給': 510, '罵': 511, '美妝': 512, '蜜': 513, '證明': 514, '需': 515, '首': 516, '鮭魚': 517, '🤣': 518, '下好': 519, '交替': 520, '伺機而動': 521, '出現': 522, '動動': 523, '回溫': 524, '外加': 525, '多加': 526, '感謝': 527, '戴口罩': 528, '手填': 529, '於是': 530, '晚上': 531, '極光': 532, '水': 533, '油脂': 534, '混合': 535, '琥珀': 536, '瓶': 537, '粉刺': 538, '臉': 539, '藍色': 540, '補': 541, '酸': 542, '上市': 543, '乾': 544, '剛': 545, '好久': 546, '快速': 547, '拍攝': 548, '櫃姐': 549, '滿': 550, '皆': 551, '聽': 552, '見': 553, '質地': 554, '輕薄': 555, '鏡頭': 556, '限量': 557, '霧面': 558, '顏色': 559, '亞歷山大': 560, '台北': 561, '嘗試': 562, '女孩': 563, '日式': 564, '毛文': 565, '海神': 566, '無從': 567, '版上': 568, '經驗': 569, '除毛': 570, '下標': 571, '中間': 572, '妝': 573, '弱': 574, '彩妝師': 575, '畫': 576, '精華': 577, '罐': 578, '置入': 579, '腦波': 580, '腦粉': 581, '著': 582, '超愛': 583, '身為': 584, '內包': 585, '推': 586, '會長': 587, '痘印': 588, '皮膚科': 589, '藻針': 590, '⭕': 591, '作品': 592, '前後': 593, '副': 594, '台南市': 595, '喜好': 596, '地點': 597, '對': 598, '小時': 599, '底色': 600, '拋棄式': 601, '拍': 602, '操作': 603, '收集': 604, '新手': 605, '時間': 606, '更多': 607, '有無': 608, '東區': 609, '條件': 610, '眉型': 611, '私訊': 612, '討論': 613, '設計': 614, '貼文': 615, '配合': 616, '針具': 617, '非': 618, '️': 619, '一間': 620, '仿': 621, '公司': 622, '包': 623, '各式各樣': 624, '好奇': 625, '小三': 626, '工廠': 627, '應該': 628, '特地': 629, '美日': 630, '貴': 631, '賺': 632, '辦法': 633, '長得': 634, '難道': 635, '口服': 636, '台南': 637, '療程': 638, '請教': 639, '進行': 640, '⬆': 641, '不屑': 642, '不懈': 643, '不過': 644, '先來': 645, '出沒': 646, '圖': 647, '張': 648, '摳': 649, '摸到': 650, '撒花': 651, '社區': 652, '終於': 653, '經過': 654, '脖子': 655, '貓貓': 656, '逼出來': 657, '防雷': 658, '打電話': 659, '時代': 660, '有放': 661, '桌上': 662, '母親節': 663, '熱情': 664, '統一': 665, '逛街': 666, '預購': 667, 'ω': 668, '一點': 669, '上班': 670, '低調': 671, '偏白': 672, '友善': 673, '妝容': 674, '挑': 675, '搞定': 676, '整個': 677, '百搭': 678, '盤': 679, '眼影盤': 680, '膚色': 681, '芋芋': 682, '配色': 683, '體積': 684, '鮮奶': 685, '黃皮': 686, '之類': 687, '保濕': 688, '先擋': 689, '凝膠': 690, '剛好': 691, '化妝水': 692, '層': 693, '差': 694, '年近': 695, '後來': 696, '懶得': 697, '效果': 698, '時候': 699, '月': 700, '狀況': 701, '用了': 702, '用完': 703, '皮膚': 704, '精華液': 705, '著用': 706, '蘆薈': 707, '中午': 708, '失敗': 709, '學會': 710, '實實在在': 711, '就塌': 712, '手殘': 713, '擁有': 714, '教學': 715, '早上': 716, '朋朋': 717, '朋朋們': 718, '濃密': 719, '獨角獸': 720, '睫毛': 721, '細軟': 722, '翹': 723, '黏貼': 724, '˘': 725, '久': 726, '乾爹': 727, '廢人': 728, '心到': 729, '收回來': 730, '文章': 731, '新年': 732, '明明': 733, '求': 734, '秋子': 735, '程序': 736, '篇': 737, '結束': 738, '總之': 739, '轉變': 740, '連假': 741, '開工': 742, '꒰': 743, '꒱': 744, '問問': 745, '意見': 746, '推銷': 747, '提升': 748, '暗沉': 749, '粗糙': 750, '膚況': 751, '蠟黃': 752, '諮詢': 753, '通路': 754, '進一步': 755, '防曬': 756, '不行': 757, '包裝': 758, '台灣': 759, '奢華': 760, '櫃': 761, '氣到': 762, '氣墊': 763, '滑': 764, '發售': 765, '發摟到': 766, '粉紅': 767, '花去': 768, '訊息': 769, '錢': 770, '關注': 771, '隔壁': 772, '雅詩蘭黛': 773, '˃': 774, '̀': 775, '́': 776, '̶': 777, '͈': 778, 'و': 779, '௰': 780, '上圖': 781, '勤勞': 782, '大理石': 783, '好辣': 784, '小粉': 785, '打給': 786, '更新': 787, '洗': 788, '異常': 789, '翻翻': 790, '賣關子': 791, '遮瑕': 792, '齁': 793, '亮顏': 794, '優': 795, '口罩': 796, '型': 797, '度優': 798, '懇請': 799, '我塑': 800, '抗汗': 801, '持久': 802, '持妝': 803, '沾': 804, '爆汗': 805, '粉餅': 806, '綠色': 807, '膚質': 808, '蘭芝': 809, '訴求': 810, '雅': 811, '霧感': 812, '顆': 813, 'о': 814, '∀': 815, '一盒': 816, '三盒': 817, '元': 818, '划算': 819, '原價': 820, '名': 821, '寶雅': 822, '心中': 823, '有出': 824, '特價': 825, '盒': 826, '約會': 827, '超級': 828, '逛': 829, '隔天': 830, '面膜': 831, 'ㄧ': 832, '原生': 833, '參差不齊': 834, '哈嚕': 835, '完妝': 836, '尤里': 837, '整理': 838, '有長': 839, '潛水': 840, '版': 841, '短': 842, '緊張': 843, '缺口': 844, '莫名': 845, '處理': 846, '頑固': 847, '並': 848, '友好': 849, '台美': 850, '同盟': 851, '民主': 852, '美台': 853, '俄羅斯': 854, '假設': 855, '出征': 856, '勢必': 857, '北約': 858, '國家': 859, '導致': 860, '戰爭': 861, '打贏': 862, '撒野': 863, '最多': 864, '歐洲': 865, '歸屬於': 866, '波蘭': 867, '烏克蘭': 868, '理由': 869, '發起': 870, '緊鄰': 871, '讓給': 872, '門口': 873, '宜蘭縣': 874, '屏東縣': 875, '桃園市': 876, '臺東縣': 877, '苗栗縣': 878, '高雄市': 879, '停完': 880, '好多': 881, '處分': 882, '電後': 883, '領': 884, '仔': 885, '全': 886, '台': 887, '死忠': 888, '沒電': 889, '缺電': 890, '肺電': 891, '陪': 892, '任用': 893, '冠上': 894, '喜劇': 895, '國內': 896, '地區': 897, '外國': 898, '大失': 899, '大臣': 900, '媒體': 901, '導演': 902, '得票率': 903, '擔任': 904, '斯基': 905, '民心': 906, '演員': 907, '澤倫': 908, '為': 909, '烏東': 910, '無論': 911, '爭議': 912, '片場': 913, '當選': 914, '總統': 915, '腐敗': 916, '貪污': 917, '連任': 918, '開戰': 919, '高': 920, '供電': 921, '台電': 922, '員工': 923, '認真': 924, '責無旁貸': 925, '停電': 926, '國民黨': 927, '好事': 928, '好像': 929, '懂': 930, '抓到': 931, '機會': 932, '民進黨': 933, '為何': 934, '看來': 935, '罄竹難書': 936, '記得': 937, '過什麼': 938, '雞蛋': 939, '骨頭': 940, '鬥': 941, '到民': 942, '去年': 943, '問題': 944, '執政黨': 945, '增加': 946, '洗地': 947, '為啥': 948, '苗苗': 949, '選票': 950, '電網': 951, '馬政府': 952, '麽': 953, '核食': 954, '照句': 955, '疏失': 956, '福食': 957, '答': 958, '範例': 959, '解答': 960, '造樣': 961, '開放': 962, '題目': 963, '一團': 964, '中華民國': 965, '台派': 966, '名字': 967, '戰成': 968, '爸爸': 969, '講': 970, '跳針': 971, '開示': 972, '位於': 973, '俄軍': 974, '將': 975, '拜登': 976, '持續': 977, '攻勢': 978, '札': 979, '核電廠': 980, '波羅': 981, '火災': 982, '熱': 983, '白宮': 984, '稍早': 985, '致電': 986, '起火': 987, '世界': 988, '全體': 989, '各國': 990, '幫忙': 991, '殉國': 992, '烏俄': 993, '物資': 994, '英勇': 995, '誒': 996, '軍民': 997, '這麼': 998, '位當': 999, '供應': 1000, '內': 1001, '全世界': 1002, '出錯': 1003, '台灣人': 1004, '喔': 1005, '年': 1006, '拜託': 1007, '接受': 1008, '發生': 1009, '社畜': 1010, '藉口': 1011, '設施': 1012, '說出去': 1013, '這麼多': 1014, '點腦': 1015, '一早': 1016, '今早': 1017, '傳出': 1018, '公布': 1019, '凌晨': 1020, '台股': 1021, '壓低': 1022, '市場': 1023, '幣圈': 1024, '想必': 1025, '指數': 1026, '數據': 1027, '百點': 1028, '股市': 1029, '與': 1030, '跳空': 1031, '金融': 1032, '開低': 1033, '震撼彈': 1034, '非農': 1035, '介意': 1036, '依舊': 1037, '做出': 1038, '別': 1039, '南部人': 1040, '反駁': 1041, '同意': 1042, '攜手': 1043, '救': 1044, '明知道': 1045, '當初': 1046, '肚子': 1047, '自得': 1048, '自業': 1049, '這下': 1050, '選擇': 1051, '面臨': 1052, '餓': 1053, '下午': 1054, '下山': 1055, '佔': 1056, '停工': 1057, '即將': 1058, '太陽': 1059, '太陽能板': 1060, '手電筒': 1061, '有讀': 1062, '發電量': 1063, '看好': 1064, '過書': 1065, '出手': 1066, '富豪': 1067, '布丁': 1068, '攻下': 1069, '死亡': 1070, '首都': 1071, '駭客': 1072, '一檔': 1073, '中共': 1074, '主動': 1075, '之餘': 1076, '停': 1077, '出擊': 1078, '包子': 1079, '原因': 1080, '台大': 1081, '大戲': 1082, '害怕': 1083, '攻不下': 1084, '普丁': 1085, '演了': 1086, '習': 1087, '自顧不暇': 1088, '落': 1089, '蔡': 1090, '處境': 1091, '裡應外合': 1092, '誘使': 1093, '龐佩奧': 1094, '全民': 1095, '吃': 1096, '大便': 1097, '幹': 1098, '沒事': 1099, '給力': 1100, '詩': 1101, '閒暇': 1102, '一昧': 1103, '事': 1104, '公投': 1105, '奇怪': 1106, '底下': 1107, '承擔': 1108, '投反': 1109, '時': 1110, '核四': 1111, '片罵': 1112, '留言區': 1113, '當然': 1114, '負一': 1115, '負載': 1116, '責任': 1117, '路': 1118, '中俄': 1119, '侵略': 1120, '別國': 1121, '包圍': 1122, '可見': 1123, '哽咽': 1124, '坦克': 1125, '妄想': 1126, '家親': 1127, '戰場': 1128, '無情': 1129, '留學生': 1130, '舔': 1131, '萬無一失': 1132, '裝甲車': 1133, '語氣': 1134, '輛': 1135, '逃離': 1136, '逃難': 1137, '過程': 1138, '遭': 1139, '不斷': 1140, '中國人': 1141, '中國黨': 1142, '位覺': 1143, '侯友宜': 1144, '反饋': 1145, '在狄卡': 1146, '執政': 1147, '年來': 1148, '底層': 1149, '廢料': 1150, '意外': 1151, '柯': 1152, '核': 1153, '滯台': 1154, '發表': 1155, '社會': 1156, '聲稱': 1157, '言論': 1158, '質疑': 1159, '青給': 1160, '韓粉': 1161, '高潮': 1162, '高興': 1163, '一線': 1164, '不可能': 1165, '中央社': 1166, '人員': 1167, '你們': 1168, '依法': 1169, '偵辦': 1170, '免洗': 1171, '出事': 1172, '動作': 1173, '台鐵': 1174, '地檢署': 1175, '大樓': 1176, '指出': 1177, '政': 1178, '暖市': 1179, '永遠': 1180, '燒掉': 1181, '犯錯': 1182, '發抖': 1183, '究責': 1184, '第一槍': 1185, '系統': 1186, '絕對': 1187, '調查': 1188, '通知': 1189, '開': 1190, '面': 1191, '年終': 1192, '建議': 1193, '減半': 1194, '獎金': 1195, '納稅': 1196, '紕漏': 1197, '自首': 1198, '解決': 1199, '辛苦': 1200, '領著': 1201, '高雄': 1202, '高雄人': 1203, '⋯': 1204, '下定決心': 1205, '債務': 1206, '加一加': 1207, '卻': 1208, '回頭': 1209, '多萬': 1210, '家人': 1211, '戒賭': 1212, '手上': 1213, '打算': 1214, '拿去': 1215, '有錢': 1216, '欠': 1217, '沈迷': 1218, '無力': 1219, '簽': 1220, '組頭': 1221, '總共': 1222, '總是': 1223, '賭球': 1224, '賺錢': 1225, '軍官': 1226, '還債': 1227, '開口': 1228, '一年': 1229, '付清': 1230, '公務員': 1231, '公教': 1232, '卡費': 1233, '土銀': 1234, '在職': 1235, '太慢': 1236, '存款': 1237, '客服': 1238, '審核': 1239, '專案': 1240, '工作天': 1241, '年收': 1242, '房貸': 1243, '收到': 1244, '股票': 1245, '萬': 1246, '負債': 1247, '貸': 1248, '交流': 1249, '住': 1250, '分配': 1251, '多久': 1252, '存': 1253, '定存': 1254, '年紀': 1255, '年輕': 1256, '懂事': 1257, '房屋': 1258, '投資': 1259, '標的': 1260, '現金': 1261, '筆': 1262, '綁': 1263, '萬元': 1264, '計畫': 1265, '運用': 1266, '開心': 1267, '高雄市區': 1268, '低迷': 1269, '合約': 1270, '情緒': 1271, '懵了': 1272, '打開': 1273, '整體': 1274, '暴漲暴跌': 1275, '熊市': 1276, '現貨': 1277, '盤面': 1278, '稱為': 1279, '著急': 1280, '處於': 1281, '行情': 1282, '試試': 1283, '買入': 1284, '震盪': 1285, '頁面': 1286, '體驗': 1287, '低': 1288, '匯差': 1289, '台幣': 1290, '外幣': 1291, '多文': 1292, '手續費': 1293, '漲時': 1294, '爬': 1295, '美元': 1296, '買賣': 1297, '買進': 1298, '賣掉': 1299, '轉回': 1300, '關於': 1301, '交易': 1302, '以圖': 1303, '來選': 1304, '前言': 1305, '勝率': 1306, '專研': 1307, '成效': 1308, '檢視': 1309, '現象': 1310, '程式': 1311, '策略': 1312, '要用': 1313, '語言': 1314, '軟體': 1315, '默默': 1316, '刷': 1317, '刷卡': 1318, '房': 1319, '扣掉': 1320, '搭配': 1321, '訂了': 1322, '訂房': 1323, '退掉': 1324, '過了': 1325, '郵局': 1326, '金融卡': 1327, '間房': 1328, '下班': 1329, '休假': 1330, '住家': 1331, '供吃': 1332, '偶爾': 1333, '加油': 1334, '包含': 1335, '外食': 1336, '女友': 1337, '家裡': 1338, '收入': 1339, '月薪': 1340, '煮飯': 1341, '玩': 1342, '生活': 1343, '純': 1344, '苦': 1345, '被動': 1346, '陪伴': 1347, '頂多': 1348, '並非': 1349, '人性': 1350, '作': 1351, '來臨': 1352, '克服': 1353, '區塊': 1354, '大師級': 1355, '實質': 1356, '弱點': 1357, '心態': 1358, '心理': 1359, '心魔': 1360, '成功': 1361, '戰勝': 1362, '技巧': 1363, '控制': 1364, '炒': 1365, '無': 1366, '獲利': 1367, '自我': 1368, '英雄': 1369, '誰': 1370, '較量': 1371, '鏈': 1372, '靈魂': 1373, '高手': 1374, '鬥爭': 1375, '位置': 1376, '區間': 1377, '午夜': 1378, '反彈': 1379, '受阻': 1380, '回調': 1381, '壓力': 1382, '小幅': 1383, '後續': 1384, '拉升': 1385, '探底': 1386, '昨晚': 1387, '期間': 1388, '比特幣': 1389, '波': 1390, '突破': 1391, '線': 1392, '試探性': 1393, '財神': 1394, '走勢': 1395, '隨後': 1396, '串聯': 1397, '交易所': 1398, '加密': 1399, '帳戶': 1400, '幣安': 1401, '收益': 1402, '本金': 1403, '火幣': 1404, '當日': 1405, '註記': 1406, '資金': 1407, '下殺': 1408, '下跌': 1409, '下路': 1410, '創逾': 1411, '升息': 1412, '原物料': 1413, '工業': 1414, '漲幅': 1415, '聯準會': 1416, '股': 1417, '費半': 1418, '週': 1419, '道瓊': 1420, '達克': 1421, '那斯': 1422, '飽爾': 1423, '高科技': 1424, '交易量': 1425, '來得及': 1426, '大戶': 1427, '投': 1428, '折': 1429, '指定': 1430, '更換': 1431, '營業員': 1432, '爭取': 1433, '百萬': 1434, '線上': 1435, '變': 1436, '辦': 1437, '通過': 1438, '初心': 1439, '利率': 1440, '家族': 1441, '手': 1442, '操作手機': 1443, '放': 1444, '爸媽': 1445, '獲得': 1446, '稍微': 1447, '規劃': 1448, '閒置': 1449, '元大': 1450, '想過': 1451, '槓桿': 1452, '油油': 1453, '粉絲': 1454, '肥宅': 1455, '錢會': 1456, '長期': 1457, '\\u200d': 1458, '其餘': 1459, '媽': 1460, '媽付': 1461, '存滿': 1462, '存錢': 1463, '存錢筒': 1464, '手機費': 1465, '月光族': 1466, '植牙': 1467, '美食': 1468, '舉手': 1469, '萬二': 1470, '薪資': 1471, '費': 1472, '購物': 1473, '開銷': 1474, '⠀': 1475, '一般人': 1476, '出國': 1477, '唸書': 1478, '大學': 1479, '小孩': 1480, '指': 1481, '教育金': 1482, '準備': 1483, '累積': 1484, '藉由': 1485, '談論': 1486, '費用': 1487, '較': 1488, '較長': 1489, '金額': 1490, '高等': 1491, '國泰': 1492, '據點': 1493, '旗下': 1494, '臨櫃': 1495, '證券戶': 1496, '辦理': 1497, '銀行': 1498, '開戶': 1499, '新光': 1500, '永豐': 1501, '版面': 1502, '證卷': 1503, '配置': 1504, '危險': 1505, '年化': 1506, '想法': 1507, '數百': 1508, '瞭解': 1509, '礦池': 1510, '聽到': 1511, '背後': 1512, '許多': 1513, '達到': 1514, '雙幣': 1515, '風險': 1516, '利嘛': 1517, '加碼': 1518, '感動': 1519, '變紅': 1520, '霸脫': 1521, '高點': 1522, '出借': 1523, '列表': 1524, '永豐金': 1525, '無法': 1526, '申請': 1527, '股息': 1528, '買過': 1529, '開發金': 1530, '分析': 1531, '卡': 1532, '在於': 1533, '大盤': 1534, '平日': 1535, '悲觀': 1536, '技術面': 1537, '提高': 1538, '時去': 1539, '有時': 1540, '目的': 1541, '看盤': 1542, '神明': 1543, '籌碼面': 1544, '落漆': 1545, '補強': 1546, '隨手': 1547, '面子': 1548, '佩服': 1549, '前情': 1550, '厲害': 1551, '外資': 1552, '大漲': 1553, '天下無敵': 1554, '巧合': 1555, '提': 1556, '族群': 1557, '答對': 1558, '美光': 1559, '臨晨': 1560, '航運': 1561, '調評': 1562, '難': 1563, '多去': 1564, '少新': 1565, '建商': 1566, '慘': 1567, '房仲': 1568, '房子': 1569, '持有': 1570, '捷運': 1571, '案例': 1572, '檯面': 1573, '蓋多': 1574, '賠': 1575, '賣房': 1576, '跌': 1577, '騙人': 1578, '上網': 1579, '保人': 1580, '在學': 1581, '工作': 1582, '查': 1583, '正職': 1584, '沒辦法': 1585, '能當': 1586, '頭期款': 1587, '選': 1588, '人問': 1589, '價值': 1590, '否認': 1591, '回事': 1592, '實話': 1593, '幣': 1594, '投機': 1595, '炒幣': 1596, '率是': 1597, '笨蛋': 1598, '行業': 1599, '中山': 1600, '事項': 1601, '動動手': 1602, '問卷': 1603, '單位': 1604, '國立': 1605, '填寫': 1606, '填答': 1607, '忙': 1608, '數位': 1609, '為例': 1610, '用戶': 1611, '研究': 1612, '要請': 1613, '誘因': 1614, '京子': 1615, '亮眼': 1616, '僅次於': 1617, '冠軍': 1618, '則': 1619, '剛典': 1620, '妻': 1621, '季八集': 1622, '安藤': 1623, '岩田': 1624, '成人級': 1625, '成績': 1626, '拿下': 1627, '排行': 1628, '改編': 1629, '政信': 1630, '旋即': 1631, '日劇': 1632, '曾經': 1633, '本劇': 1634, '殭屍': 1635, '漫畫': 1636, '熱播': 1637, '熱銷': 1638, '登上': 1639, '篠原涼子': 1640, '金魚': 1641, '長谷川': 1642, '開播': 1643, '黑澤': 1644, '不聞不問': 1645, '人群': 1646, '孩子': 1647, '幫助': 1648, '新': 1649, '法官': 1650, '注視': 1651, '海報': 1652, '淚目': 1653, '漠然': 1654, '理念': 1655, '發布': 1656, '立場': 1657, '羅瑾熙': 1658, '部長': 1659, '青少年': 1660, '上線': 1661, '十月': 1662, '底': 1663, '獵人': 1664, '⌚': 1665, '上架': 1666, '傳記': 1667, '冒險片': 1668, '劇情': 1669, '旗幟': 1670, '極地': 1671, '歷史': 1672, '海賊': 1673, '滿愛': 1674, '議題': 1675, '這禮拜': 1676, '電影': 1677, '霜雪': 1678, '鬼怪': 1679, '🥇': 1680, '🥈': 1681, '🥉': 1682, '原來': 1683, '搞': 1684, '笑死': 1685, '轉': 1686, '主角': 1687, '主題曲': 1688, '作為': 1689, '劇中': 1690, '告別': 1691, '崔宥莉': 1692, '心情': 1693, '心愛': 1694, '悲傷': 1695, '慢慢': 1696, '描述': 1697, '收聽': 1698, '有限': 1699, '歌手': 1700, '歌曲': 1701, '燦榮': 1702, '生命': 1703, '田美': 1704, '發行': 1705, '網站': 1706, '鄭': 1707, '音源': 1708, '一集': 1709, '三大': 1710, '了本': 1711, '全球': 1712, '厭惡': 1713, '口碑': 1714, '少年': 1715, '少年犯': 1716, '法庭': 1717, '為止': 1718, '特色': 1719, '犯罪': 1720, '發酵': 1721, '至極': 1722, '英語': 1723, '觀眾': 1724, '說明': 1725, '躍升': 1726, '金惠秀': 1727, '開宗明義': 1728, '韓國': 1729, '題材': 1730, '飾演': 1731, '首度': 1732, '不適': 1733, '佐': 1734, '律師': 1735, '文森': 1736, '殘忍': 1737, '特別': 1738, '結局': 1739, '臥床': 1740, '身體': 1741, '遊戲': 1742, '驅魔': 1743, '魷魚': 1744, '麵館': 1745, '黑道': 1746, '二十五二十一': 1747, '保重': 1748, '劇組': 1749, '南柱赫': 1750, '影響': 1751, '恢復': 1752, '播出': 1753, '新冠': 1754, '日後': 1755, '檢測': 1756, '泰梨': 1757, '為事': 1758, '確診': 1759, '肺炎': 1760, '苞娜': 1761, '週末': 1762, '金泰梨': 1763, '陰性': 1764, '隔離': 1765, '下一集': 1766, '手點': 1767, '播放': 1768, '智慧': 1769, '求助': 1770, '自動': 1771, '超過': 1772, '電腦': 1773, '電視': 1774, '受': 1775, '圖用': 1776, '失望': 1777, '好評': 1778, '密碼': 1779, '廁所': 1780, '愛上': 1781, '男女': 1782, '痛苦': 1783, '相親': 1784, '看完': 1785, '社內': 1786, '等待': 1787, '集': 1788, '集換': 1789, '順': 1790, '馬上': 1791, '劇': 1792, '合作': 1793, '審判': 1794, '律政': 1795, '日': 1796, '洪': 1797, '私生活': 1798, '編劇': 1799, '金玟錫': 1800, '鍾燦': 1801, '韓語': 1802, '년': 1803, '소': 1804, '심': 1805, '판': 1806, '一月份': 1807, '中華': 1808, '二月': 1809, '付': 1810, '份': 1811, '信用卡': 1812, '帳單': 1813, '月租費': 1814, '盜刷': 1815, '肯定': 1816, '自從': 1817, '計費': 1818, '週期': 1819, '電信': 1820, '季播': 1821, '黑鏡': 1822, '下載': 1823, '人用': 1824, '字幕': 1825, '字體': 1826, '應用': 1827, '網頁版': 1828, '同時': 1829, '觀看': 1830, '作畫': 1831, '健': 1832, '出水': 1833, '判官': 1834, '動畫': 1835, '原創': 1836, '國王': 1837, '小圓': 1838, '小畑': 1839, '巨人': 1840, '心靈': 1841, '排名': 1842, '東京': 1843, '淹水': 1844, '潮到': 1845, '虛淵玄': 1846, '製作': 1847, '角色': 1848, '進擊': 1849, '魔法': 1850, '具': 1851, '原': 1852, '娑婆': 1853, '封': 1854, '專輯': 1855, '小將': 1856, '景伊': 1857, '柔美': 1858, '毒樓': 1859, '法院': 1860, '泰成': 1861, '監製': 1862, '祭司': 1863, '紙牌屋': 1864, '細胞': 1865, '聲帶': 1866, '訶': 1867, '那一天': 1868, '配樂': 1869, '金': 1870, '雪降': 1871, '電視劇': 1872, '音樂': 1873, '黎明': 1874, '黑': 1875, '些能': 1876, '人身': 1877, '壞': 1878, '多數': 1879, '家庭': 1880, '少數': 1881, '必須': 1882, '成了': 1883, '教育出': 1884, '歸咎於': 1885, '規避': 1886, '過錯': 1887, '雖說': 1888, '名稱': 1889, '嗚': 1890, '嗚嗚': 1891, '城市': 1892, '好好笑': 1893, '捨不得': 1894, '期待': 1895, '翻拍': 1896, '聽說': 1897, '上來': 1898, '下手': 1899, '不出': 1900, '大神': 1901, '女子': 1902, '小說': 1903, '所以然': 1904, '查了': 1905, '查查': 1906, '版本': 1907, '興趣': 1908, '順序': 1909, '🤨': 1910, '一部': 1911, '保健': 1912, '傑克': 1913, '戲': 1914, '拿走': 1915, '提問': 1916, '當天': 1917, '腎上腺素': 1918, '自殺': 1919, '賽門': 1920, '小黑': 1921, '查不到': 1922, '白癡': 1923, '能用': 1924, '蘋果': 1925, '訂閱': 1926, '主演': 1927, '取自': 1928, '圍繞': 1929, '季': 1930, '季度': 1931, '極度': 1932, '秀': 1933, '節目': 1934, '簡介': 1935, '處決': 1936, '討厭': 1937, '認同': 1938, '講述': 1939, '金憓': 1940, '金武烈': 1941, '韓劇': 1942, '人權': 1943, '口號': 1944, '和平': 1945, '回應': 1946, '失': 1947, '屬於': 1948, '廣場': 1949, '抱有': 1950, '換來': 1951, '政府': 1952, '暴力': 1953, '最初': 1954, '未來': 1955, '演變成': 1956, '獨立': 1957, '當時': 1958, '發展': 1959, '百': 1960, '萬人': 1961, '血腥': 1962, '見證': 1963, '運動': 1964, '革命': 1965, '齊聚': 1966, '封面': 1967, '演唱': 1968, '熱烈': 1969, '男團': 1970, 'Д': 1971, 'ຶ': 1972, '不長': 1973, '分局': 1974, '影集': 1975, '推爆': 1976, '耐看': 1977, '荒謬': 1978, '追': 1979, '前三集': 1980, '寫實': 1981, '悶': 1982, '服裝': 1983, '板發': 1984, '格陵蘭': 1985, '正傳': 1986, '段': 1987, '海豹': 1988, '渡海': 1989, '無聊': 1990, '皮': 1991, '細緻': 1992, '老實': 1993, '考據': 1994, '船': 1995, '部': 1996, '部位': 1997, '共鳴': 1998, '大推': 1999, '性格': 2000, '提到': 2001, '海嘯': 2002, '疫情': 2003, '聲音': 2004, '討喜': 2005, '讚美': 2006, '重雷': 2007, '開頭': 2008}\n"
     ]
    }
   ],
   "source": [
    "texts = sample_df[\"toks\"].to_list()\n",
    "dictionary = corpora.Dictionary(texts)\n",
    "print(dictionary.token2id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "_W7Pha8ByNbt",
    "outputId": "dcf20097-df26-435c-b63f-5c1ccb3357ac"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dictionary(2009 unique tokens: ['位', '個', '填', '希望', '幫']...)\n"
     ]
    }
   ],
   "source": [
    "print(dictionary)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "id": "71i19CBfq4x-"
   },
   "outputs": [],
   "source": [
    "corpus = [dictionary.doc2bow(text) for text in texts]\n",
    "tfidf = models.TfidfModel(corpus)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rSneNC-hzmzM",
    "outputId": "5a560b38-d654-49cf-9ec0-6a56b2847f46"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "140"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dictionary.num_docs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3uO1g_ebzy-v"
   },
   "source": [
    "- You can change the value of `docid` to any number between 0 and 139."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "UjYxfmW4zzU-",
    "outputId": "320e8c8a-5524-40dc-c56e-fb6b17c70a86"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Raw sample text:\n",
      "一個是等烏克蘭首都被攻下，一個是等布丁死亡，有俄羅斯自家富豪出手了，BTW俄羅斯自家的駭客也在幫忙打布丁\n",
      "Tokens of sample text:\n",
      "['個', '烏克蘭', '首都', '攻下', '個', '布丁', '死亡', '俄羅斯', '富豪', '出手', '俄羅斯', '駭客', '幫忙', '布丁']\n"
     ]
    }
   ],
   "source": [
    "docid = 73\n",
    "raw_text_in_corpus = sample_df.loc[docid,'excerpt']\n",
    "tokens_in_raw_text = sample_df.loc[docid,'toks']\n",
    "print(\"Raw sample text:\")\n",
    "print(raw_text_in_corpus)\n",
    "print(\"Tokens of sample text:\")\n",
    "print(tokens_in_raw_text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "O8fbzR4jz_Wk"
   },
   "source": [
    "- A gensim corpus stores a mapping between word IDs and their frequencies in a document."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "uLP0p9hT8TaV",
    "outputId": "3e9cc8fc-3a5a-4111-b8ca-e70062ba56ed"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(1, 2),\n",
       " (854, 2),\n",
       " (868, 1),\n",
       " (991, 1),\n",
       " (1066, 1),\n",
       " (1067, 1),\n",
       " (1068, 2),\n",
       " (1069, 1),\n",
       " (1070, 1),\n",
       " (1071, 1),\n",
       " (1072, 1)]"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corpus[docid]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "xSiB1-wz8dSD",
    "outputId": "5c2de1e0-1062-4880-f25a-6f8281c8f0ed"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "個 >>> occurs 2 times(s) in this doc\n",
      "俄羅斯 >>> occurs 2 times(s) in this doc\n",
      "烏克蘭 >>> occurs 1 times(s) in this doc\n",
      "幫忙 >>> occurs 1 times(s) in this doc\n",
      "出手 >>> occurs 1 times(s) in this doc\n",
      "富豪 >>> occurs 1 times(s) in this doc\n",
      "布丁 >>> occurs 2 times(s) in this doc\n",
      "攻下 >>> occurs 1 times(s) in this doc\n",
      "死亡 >>> occurs 1 times(s) in this doc\n",
      "首都 >>> occurs 1 times(s) in this doc\n",
      "駭客 >>> occurs 1 times(s) in this doc\n"
     ]
    }
   ],
   "source": [
    "for id, freq in corpus[docid]:\n",
    "    print(f\"{dictionary[id]} >>> occurs {freq} times(s) in this doc\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xBjIIve409I3"
   },
   "source": [
    "- A gensim model stores a mapping between word IDs and their scores as measured by that model based on the corpus."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "uTZn9yar8zUk",
    "outputId": "656dae9d-cd49-4378-c31b-9181be99560e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(1, 0.15147118274062674),\n",
       " (854, 0.3716399607099024),\n",
       " (868, 0.15304212827789748),\n",
       " (991, 0.23691682188324237),\n",
       " (1066, 0.2755700903370344),\n",
       " (1067, 0.2755700903370344),\n",
       " (1068, 0.5511401806740688),\n",
       " (1069, 0.2755700903370344),\n",
       " (1070, 0.2755700903370344),\n",
       " (1071, 0.2755700903370344),\n",
       " (1072, 0.2755700903370344)]"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tfidf[corpus[docid]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "u_P0sAfC8_a3",
    "outputId": "e2adbc38-ca04-48b3-a226-3dcda564112c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "個 >>> whose tfidf is 0.15147118274062674 in the corpus\n",
      "俄羅斯 >>> whose tfidf is 0.3716399607099024 in the corpus\n",
      "烏克蘭 >>> whose tfidf is 0.15304212827789748 in the corpus\n",
      "幫忙 >>> whose tfidf is 0.23691682188324237 in the corpus\n",
      "出手 >>> whose tfidf is 0.2755700903370344 in the corpus\n",
      "富豪 >>> whose tfidf is 0.2755700903370344 in the corpus\n",
      "布丁 >>> whose tfidf is 0.5511401806740688 in the corpus\n",
      "攻下 >>> whose tfidf is 0.2755700903370344 in the corpus\n",
      "死亡 >>> whose tfidf is 0.2755700903370344 in the corpus\n",
      "首都 >>> whose tfidf is 0.2755700903370344 in the corpus\n",
      "駭客 >>> whose tfidf is 0.2755700903370344 in the corpus\n"
     ]
    }
   ],
   "source": [
    "for id, score in tfidf[corpus[docid]]:\n",
    "    print(f\"{dictionary[id]} >>> whose tfidf is {score} in the corpus\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lVeebTHIq_PM"
   },
   "source": [
    "- With the TF-IDF model, we can now convert any word in the corpus to a number.\n",
    "- To convert a document into number, we'll need the Document-Term Matrix (DTM).\n",
    "![](https://rlads2021.github.io/LabBook/assets/img/dtm.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "id": "-BfAFwKbmi8N"
   },
   "outputs": [],
   "source": [
    "def vectorize_doc(docid):\n",
    "    id_tfidf = tfidf[corpus[docid]]\n",
    "    vector = {dictionary[id]: tfidf for id, tfidf in id_tfidf}\n",
    "    return vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "id": "xmFwewPepygj"
   },
   "outputs": [],
   "source": [
    "dtm_data = [vectorize_doc(docid) for docid in sample_df.index]\n",
    "dtm_df = pd.DataFrame(dtm_data)\n",
    "dtm_df.fillna(0, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 487
    },
    "id": "h6jixS7yqhRL",
    "outputId": "e5f2479f-f1ed-4e98-d2fd-ffedf66636b5"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-e19b54f2-ae25-4d87-bcdd-84d9e7f20cf6\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>位</th>\n",
       "      <th>個</th>\n",
       "      <th>填</th>\n",
       "      <th>希望</th>\n",
       "      <th>幫</th>\n",
       "      <th>秒</th>\n",
       "      <th>花</th>\n",
       "      <th>囝仔</th>\n",
       "      <th>圖是</th>\n",
       "      <th>套</th>\n",
       "      <th>...</th>\n",
       "      <th>大推</th>\n",
       "      <th>性格</th>\n",
       "      <th>提到</th>\n",
       "      <th>海嘯</th>\n",
       "      <th>疫情</th>\n",
       "      <th>聲音</th>\n",
       "      <th>討喜</th>\n",
       "      <th>讚美</th>\n",
       "      <th>重雷</th>\n",
       "      <th>開頭</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.281965</td>\n",
       "      <td>0.155875</td>\n",
       "      <td>0.487609</td>\n",
       "      <td>0.328501</td>\n",
       "      <td>0.343826</td>\n",
       "      <td>0.487609</td>\n",
       "      <td>0.441073</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.292414</td>\n",
       "      <td>0.292414</td>\n",
       "      <td>0.227406</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.127547</td>\n",
       "      <td>0.070510</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>135</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>137</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>138</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>139</th>\n",
       "      <td>0.112257</td>\n",
       "      <td>0.062058</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "      <td>0.225801</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>140 rows × 2009 columns</p>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-e19b54f2-ae25-4d87-bcdd-84d9e7f20cf6')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-e19b54f2-ae25-4d87-bcdd-84d9e7f20cf6 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-e19b54f2-ae25-4d87-bcdd-84d9e7f20cf6');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "            位         個         填        希望         幫         秒         花  \\\n",
       "0    0.281965  0.155875  0.487609  0.328501  0.343826  0.487609  0.441073   \n",
       "1    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "2    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "3    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "4    0.127547  0.070510  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "..        ...       ...       ...       ...       ...       ...       ...   \n",
       "135  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "136  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "137  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "138  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "139  0.112257  0.062058  0.000000  0.000000  0.000000  0.000000  0.000000   \n",
       "\n",
       "           囝仔        圖是         套  ...        大推        性格        提到  \\\n",
       "0    0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "1    0.292414  0.292414  0.227406  ...  0.000000  0.000000  0.000000   \n",
       "2    0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "3    0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "4    0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "..        ...       ...       ...  ...       ...       ...       ...   \n",
       "135  0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "136  0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "137  0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "138  0.000000  0.000000  0.000000  ...  0.000000  0.000000  0.000000   \n",
       "139  0.000000  0.000000  0.000000  ...  0.225801  0.225801  0.225801   \n",
       "\n",
       "           海嘯        疫情        聲音        討喜        讚美        重雷        開頭  \n",
       "0    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "1    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "2    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "3    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "4    0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "..        ...       ...       ...       ...       ...       ...       ...  \n",
       "135  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "136  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "137  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "138  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  0.000000  \n",
       "139  0.225801  0.225801  0.225801  0.225801  0.225801  0.225801  0.225801  \n",
       "\n",
       "[140 rows x 2009 columns]"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dtm_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IEeoFdH_gPyg"
   },
   "source": [
    "- Each document can now be prepresented by a vector in 2009 dimensions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "4g6xu0yYtY6E",
    "outputId": "cb9b86f3-a604-4f70-a036-e909aaa65646"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Text:\n",
      "一個是等烏克蘭首都被攻下，一個是等布丁死亡，有俄羅斯自家富豪出手了，BTW俄羅斯自家的駭客也在幫忙打布丁\n",
      "Vector:\n",
      "位     0.000000\n",
      "個     0.151471\n",
      "填     0.000000\n",
      "希望    0.000000\n",
      "幫     0.000000\n",
      "        ...   \n",
      "聲音    0.000000\n",
      "討喜    0.000000\n",
      "讚美    0.000000\n",
      "重雷    0.000000\n",
      "開頭    0.000000\n",
      "Name: 73, Length: 2009, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "text = sample_df.loc[docid, \"excerpt\"]\n",
    "vec = dtm_df.loc[docid]\n",
    "print(f\"Text:\\n{text}\")\n",
    "print(f\"Vector:\\n{vec}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "HF--37xk_ZQd"
   },
   "source": [
    "- Disadvantages of TF-IDF\n",
    " - A document is represented by a vector in N dimensions (where N==size of the vocab). \n",
    " - The values of many dimensions in a vector are 0 (called `sparse vectors`).  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CyoMu0HZtHm1",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### PCA (Principal Component Analysis)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0qiXfYbs3VSF"
   },
   "source": [
    "- PCA, a technique for reducing dimensions\n",
    "\n",
    "![](https://miro.medium.com/max/1400/1*mgncZaKaVx9U6OCQu_m8Bg.jpeg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "X__r24m-v0e7",
    "outputId": "33b056b6-51d7-4ed6-b197-a2a7638ea1d5"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.7/dist-packages/distributed/config.py:20: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.\n",
      "  defaults = yaml.load(f)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "import numpy as np\n",
    "import plotly.express as px"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "PrZSaovIfOlY"
   },
   "source": [
    "- With PCA, we can easily reduce the dimension of a document vector from 2009 to 20.\n",
    "- After the reduction, a sparse vector becomes a dense one."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "oxma-Tcfv2kB",
    "outputId": "b6215f59-0ca4-449a-8ff8-35e3cff2b3ca"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.09593937,  0.08866023,  0.07769523, -0.09343106, -0.20431264,\n",
       "        0.139582  , -0.29602169, -0.02982684, -0.24764987, -0.02938505,\n",
       "        0.04872477,  0.04076477,  0.04735796,  0.0764895 , -0.13701334,\n",
       "        0.00522563,  0.00850476,  0.03513411,  0.05439893,  0.07387785])"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pca = PCA(n_components=20)\n",
    "transformed_data = pca.fit_transform(dtm_df)\n",
    "transformed_data[docid]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rHjLsdSZyX0v",
    "outputId": "1b326075-8b17-4477-8ae1-f751d3deb249"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: plotly in /usr/local/lib/python3.7/dist-packages (5.6.0)\n",
      "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from plotly) (1.15.0)\n",
      "Requirement already satisfied: tenacity>=6.2.0 in /usr/local/lib/python3.7/dist-packages (from plotly) (8.0.1)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "!pip install --upgrade plotly"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "PIpa8P9Ty3FZ",
    "outputId": "2d02b395-8535-4149-cb70-5d7afde2431c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: pyyaml==5.4.1 in /usr/local/lib/python3.7/dist-packages (5.4.1)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "!pip install pyyaml==5.4.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 542
    },
    "id": "mtNUC96X8Tfq",
    "outputId": "0bd4c718-7e86-4d38-f1c8-37e1d497b8d5"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<html>\n",
       "<head><meta charset=\"utf-8\" /></head>\n",
       "<body>\n",
       "    <div>            <script src=\"https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/MathJax.js?config=TeX-AMS-MML_SVG\"></script><script type=\"text/javascript\">if (window.MathJax) {MathJax.Hub.Config({SVG: {font: \"STIX-Web\"}});}</script>                <script type=\"text/javascript\">window.PlotlyConfig = {MathJaxConfig: 'local'};</script>\n",
       "        <script src=\"https://cdn.plot.ly/plotly-2.9.0.min.js\"></script>                <div id=\"0acf6b87-4287-4d55-a41e-a1d46873cf8c\" class=\"plotly-graph-div\" style=\"height:525px; width:100%;\"></div>            <script type=\"text/javascript\">                                    window.PLOTLYENV=window.PLOTLYENV || {};                                    if (document.getElementById(\"0acf6b87-4287-4d55-a41e-a1d46873cf8c\")) {                    Plotly.newPlot(                        \"0acf6b87-4287-4d55-a41e-a1d46873cf8c\",                        [{\"hovertemplate\":\"# Components=%{x}<br>Explained Variance %=%{y}<extra></extra>\",\"legendgroup\":\"\",\"line\":{\"color\":\"#636efa\"},\"marker\":{\"symbol\":\"circle\"},\"mode\":\"lines\",\"name\":\"\",\"orientation\":\"v\",\"showlegend\":false,\"stackgroup\":\"1\",\"x\":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120],\"xaxis\":\"x\",\"y\":[1.3998883760791878,2.5766640790009476,3.736050940965008,4.8517040845458865,5.946068210494003,7.028399340563445,8.09606892721001,9.134555234942436,10.169826414072427,11.181761624708603,12.180819742215249,13.169264380470466,14.152786911423467,15.13098712710386,16.092612564091787,17.037825672061278,17.973515734204618,18.898784062619598,19.81214062808265,20.710871491682507,21.60628416082863,22.49178801004707,23.372976092554985,24.251775415531007,25.121931654372595,25.990214512165238,26.845750943965143,27.700499466678387,28.5408467925573,29.37831805038213,30.210271292917643,31.036965075378504,31.863401043278948,32.682328001395724,33.49579383666399,34.30111507763878,35.10579066929896,35.90398156679666,36.698344731696174,37.491626731297316,38.28274493585411,39.06552003790608,39.84490536949502,40.622772494027906,41.39211837400501,42.158889349595526,42.924614961674166,43.687916860042854,44.44617150079803,45.20058614177711,45.95214838260733,46.699939581632684,47.44478175381482,48.18846352776527,48.92825465582811,49.66508760751557,50.400752212389996,51.130311813015375,51.85766634232949,52.5822991688106,53.30368585779507,54.0244594879101,54.73955209055822,55.4539485795223,56.164159426049714,56.87172457356198,57.575913663648514,58.27915027929385,58.98203451233645,59.681149261039025,60.37511159976564,61.067808323924886,61.756909585136896,62.44376575010904,63.125686603447896,63.80653054845251,64.48295872613599,65.15838552212932,65.83115865337084,66.5011552641583,67.16817121486636,67.83409872979064,68.4959412677631,69.15612019049576,69.81313022017272,70.46905041405377,71.12164063274058,71.77168458400212,72.41989810526785,73.06395295839066,73.70642969015648,74.34544979773997,74.98124862707337,75.61662862231037,76.25010060290546,76.88190395674073,77.51194820591631,78.1383335459109,78.76181334166534,79.38006558674692,79.99755360933172,80.60779815515559,81.21616780255007,81.82088810481913,82.42092719067317,83.01812905999545,83.61114985628488,84.20334473225896,84.78996415582466,85.37382830076795,85.95400411389096,86.53002512099116,87.10488223513285,87.67744955335894,88.24514300598719,88.81148849618694,89.37324229548028,89.92596913872985,90.47469696502016,91.02216151145632],\"yaxis\":\"y\",\"type\":\"scatter\"}],                        {\"template\":{\"data\":{\"bar\":[{\"error_x\":{\"color\":\"#2a3f5f\"},\"error_y\":{\"color\":\"#2a3f5f\"},\"marker\":{\"line\":{\"color\":\"#E5ECF6\",\"width\":0.5},\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"bar\"}],\"barpolar\":[{\"marker\":{\"line\":{\"color\":\"#E5ECF6\",\"width\":0.5},\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"barpolar\"}],\"carpet\":[{\"aaxis\":{\"endlinecolor\":\"#2a3f5f\",\"gridcolor\":\"white\",\"linecolor\":\"white\",\"minorgridcolor\":\"white\",\"startlinecolor\":\"#2a3f5f\"},\"baxis\":{\"endlinecolor\":\"#2a3f5f\",\"gridcolor\":\"white\",\"linecolor\":\"white\",\"minorgridcolor\":\"white\",\"startlinecolor\":\"#2a3f5f\"},\"type\":\"carpet\"}],\"choropleth\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"choropleth\"}],\"contour\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"contour\"}],\"contourcarpet\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"contourcarpet\"}],\"heatmap\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"heatmap\"}],\"heatmapgl\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"heatmapgl\"}],\"histogram\":[{\"marker\":{\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"histogram\"}],\"histogram2d\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"histogram2d\"}],\"histogram2dcontour\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"histogram2dcontour\"}],\"mesh3d\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"mesh3d\"}],\"parcoords\":[{\"line\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"parcoords\"}],\"pie\":[{\"automargin\":true,\"type\":\"pie\"}],\"scatter\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatter\"}],\"scatter3d\":[{\"line\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatter3d\"}],\"scattercarpet\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattercarpet\"}],\"scattergeo\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattergeo\"}],\"scattergl\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattergl\"}],\"scattermapbox\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattermapbox\"}],\"scatterpolar\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterpolar\"}],\"scatterpolargl\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterpolargl\"}],\"scatterternary\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterternary\"}],\"surface\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"surface\"}],\"table\":[{\"cells\":{\"fill\":{\"color\":\"#EBF0F8\"},\"line\":{\"color\":\"white\"}},\"header\":{\"fill\":{\"color\":\"#C8D4E3\"},\"line\":{\"color\":\"white\"}},\"type\":\"table\"}]},\"layout\":{\"annotationdefaults\":{\"arrowcolor\":\"#2a3f5f\",\"arrowhead\":0,\"arrowwidth\":1},\"autotypenumbers\":\"strict\",\"coloraxis\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"colorscale\":{\"diverging\":[[0,\"#8e0152\"],[0.1,\"#c51b7d\"],[0.2,\"#de77ae\"],[0.3,\"#f1b6da\"],[0.4,\"#fde0ef\"],[0.5,\"#f7f7f7\"],[0.6,\"#e6f5d0\"],[0.7,\"#b8e186\"],[0.8,\"#7fbc41\"],[0.9,\"#4d9221\"],[1,\"#276419\"]],\"sequential\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"sequentialminus\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]]},\"colorway\":[\"#636efa\",\"#EF553B\",\"#00cc96\",\"#ab63fa\",\"#FFA15A\",\"#19d3f3\",\"#FF6692\",\"#B6E880\",\"#FF97FF\",\"#FECB52\"],\"font\":{\"color\":\"#2a3f5f\"},\"geo\":{\"bgcolor\":\"white\",\"lakecolor\":\"white\",\"landcolor\":\"#E5ECF6\",\"showlakes\":true,\"showland\":true,\"subunitcolor\":\"white\"},\"hoverlabel\":{\"align\":\"left\"},\"hovermode\":\"closest\",\"mapbox\":{\"style\":\"light\"},\"paper_bgcolor\":\"white\",\"plot_bgcolor\":\"#E5ECF6\",\"polar\":{\"angularaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"bgcolor\":\"#E5ECF6\",\"radialaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"}},\"scene\":{\"xaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"},\"yaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"},\"zaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"}},\"shapedefaults\":{\"line\":{\"color\":\"#2a3f5f\"}},\"ternary\":{\"aaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"baxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"bgcolor\":\"#E5ECF6\",\"caxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"}},\"title\":{\"x\":0.05},\"xaxis\":{\"automargin\":true,\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\",\"title\":{\"standoff\":15},\"zerolinecolor\":\"white\",\"zerolinewidth\":2},\"yaxis\":{\"automargin\":true,\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\",\"title\":{\"standoff\":15},\"zerolinecolor\":\"white\",\"zerolinewidth\":2}}},\"xaxis\":{\"anchor\":\"y\",\"domain\":[0.0,1.0],\"title\":{\"text\":\"# Components\"}},\"yaxis\":{\"anchor\":\"x\",\"domain\":[0.0,1.0],\"title\":{\"text\":\"Explained Variance %\"}},\"legend\":{\"tracegroupgap\":0},\"margin\":{\"t\":60}},                        {\"responsive\": true}                    ).then(function(){\n",
       "                            \n",
       "var gd = document.getElementById('0acf6b87-4287-4d55-a41e-a1d46873cf8c');\n",
       "var x = new MutationObserver(function (mutations, observer) {{\n",
       "        var display = window.getComputedStyle(gd).display;\n",
       "        if (!display || display === 'none') {{\n",
       "            console.log([gd, 'removed!']);\n",
       "            Plotly.purge(gd);\n",
       "            observer.disconnect();\n",
       "        }}\n",
       "}});\n",
       "\n",
       "// Listen for the removal of the full notebook cells\n",
       "var notebookContainer = gd.closest('#notebook-container');\n",
       "if (notebookContainer) {{\n",
       "    x.observe(notebookContainer, {childList: true});\n",
       "}}\n",
       "\n",
       "// Listen for the clearing of the current output cell\n",
       "var outputEl = gd.closest('.output');\n",
       "if (outputEl) {{\n",
       "    x.observe(outputEl, {childList: true});\n",
       "}}\n",
       "\n",
       "                        })                };                            </script>        </div>\n",
       "</body>\n",
       "</html>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pca = PCA(n_components=120)\n",
    "pca.fit(dtm_df)\n",
    "exp_var_cumul = np.cumsum(pca.explained_variance_ratio_ * 100)\n",
    "px.area(\n",
    "    x=range(1, exp_var_cumul.shape[0] + 1),\n",
    "    y=exp_var_cumul,\n",
    "    labels={\"x\": \"# Components\", \"y\": \"Explained Variance %\"}\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 542
    },
    "id": "wlwFYh90xoC6",
    "outputId": "c67e451d-e567-4e23-bdbd-6331f1f8b900"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<html>\n",
       "<head><meta charset=\"utf-8\" /></head>\n",
       "<body>\n",
       "    <div>            <script src=\"https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/MathJax.js?config=TeX-AMS-MML_SVG\"></script><script type=\"text/javascript\">if (window.MathJax) {MathJax.Hub.Config({SVG: {font: \"STIX-Web\"}});}</script>                <script type=\"text/javascript\">window.PlotlyConfig = {MathJaxConfig: 'local'};</script>\n",
       "        <script src=\"https://cdn.plot.ly/plotly-2.9.0.min.js\"></script>                <div id=\"75fd6a97-d7d0-47cb-ba2e-434d4d74fd03\" class=\"plotly-graph-div\" style=\"height:525px; width:100%;\"></div>            <script type=\"text/javascript\">                                    window.PLOTLYENV=window.PLOTLYENV || {};                                    if (document.getElementById(\"75fd6a97-d7d0-47cb-ba2e-434d4d74fd03\")) {                    Plotly.newPlot(                        \"75fd6a97-d7d0-47cb-ba2e-434d4d74fd03\",                        [{\"hovertemplate\":\"color=dressup<br>PC 1=%{x}<br>PC 2=%{y}<br>PC 3=%{z}<extra></extra>\",\"legendgroup\":\"dressup\",\"marker\":{\"color\":\"#636efa\",\"symbol\":\"circle\"},\"mode\":\"markers\",\"name\":\"dressup\",\"scene\":\"scene\",\"showlegend\":true,\"x\":[0.016415187820039763,0.20787437054520977,-0.03085002482021088,0.12985681546075556,0.08408900451578791,0.1569029207485042,0.02629379701580087,0.17628681969830318,-0.02687056226394877,0.20827385028221157,0.38355852090800135,0.10322572252305705,0.44050053716121873,0.07982671546911534,-0.02516338142842242,0.314564340488648,0.03503020638456636,0.26878285661724677,0.2089421992978265,0.07285435809349822,-0.0454776577804687,0.000984269790769194,0.31326747682943534,0.1659069428624633,0.26443321167569966,0.2722369276150871,0.44426888004499815,0.1939270774121603],\"y\":[-0.0585886321211156,-0.03147979135744821,-0.07136322077010758,-0.04732439048956831,-0.10227687250251952,0.12712053079735788,-0.052370838276752485,-0.012150618480143026,-0.03624774569578219,-0.1803728664922133,0.08399585060610602,-0.04843495861656297,-0.01733302323434298,-0.055666840665665594,-0.03798874200884608,-0.18817153117970944,-0.07648672850022697,-0.004166480535367491,-0.14477272587645973,-0.048337910048157205,-0.023735315319070718,-0.07512751340760013,-0.062294516843776016,-0.17830294714963374,-0.22104078210260564,0.6320417828827081,0.5480010639317086,0.1069293866070018],\"z\":[0.115224441842557,0.06518532332706531,-0.06665090302424773,-0.05117102781758114,0.005587094545124839,0.016988500142288674,-0.0069390434017693485,0.04486333310523944,0.06966810577728418,0.18632285901236514,-0.08534671851727207,0.24551420487267622,-0.08531094807620408,0.1201948322004889,-0.032938210446415775,-0.037887250415125286,0.057532550169693386,0.00602217918530207,-0.10880396258884699,-0.21412372896544377,0.017422207853990054,-0.029293815554980618,-0.09877300697776885,-0.24134793757343206,0.04163372323972892,0.07163241785104955,-0.053705325125767686,0.030321656588704584],\"type\":\"scatter3d\"},{\"hovertemplate\":\"color=makeup<br>PC 1=%{x}<br>PC 2=%{y}<br>PC 3=%{z}<extra></extra>\",\"legendgroup\":\"makeup\",\"marker\":{\"color\":\"#EF553B\",\"symbol\":\"circle\"},\"mode\":\"markers\",\"name\":\"makeup\",\"scene\":\"scene\",\"showlegend\":true,\"x\":[0.06423248635426387,0.020785287704577698,-0.005338552412547427,0.014233363333140691,-0.046281707714070065,-0.01996016224061321,-0.035281145812167784,0.015910579505805524,0.0017030692456568794,0.0011304599577564747,0.13748688846860174,0.07168687835252001,0.05425548559353608,-0.05868131671497716,-0.02597694691946185,0.04638553624855924,-0.016053998014581777,0.08976266099074726,0.04955062056443657,0.05964146668756988,-0.07663430443012204,0.0022612362737413887,0.07264158259673362,-0.05470628914814989,-0.02393285766285147,0.012086899466081743,-9.165628753628904e-05,-0.09841466397681153],\"y\":[-0.08613706519161031,-0.06519518555518558,-0.06113878633686816,0.07071444622908984,-0.03969845985563677,0.01851743006255451,-0.059249865724289336,-0.015437376861759576,-0.09894107497167913,-0.11773792412840522,-0.09467173084593823,-0.11929644035630159,-0.20045085263359366,-0.030988959953212315,0.060542815924112696,-0.09883157241999155,-0.13720425969953673,0.12411652206584917,-0.07463935569534128,0.08328867331711994,-0.05584948202195035,-0.046089228286814776,-0.052357685354867144,0.03128356736307408,-0.12531239984428885,-0.054538317673510445,-0.05444577828130407,-0.05409594093062376],\"z\":[-0.2213416075536124,-0.13723267431660396,-0.1157421310937894,0.05398419868319112,-0.06925941070265434,0.030691194085246645,-0.017739453354958197,-0.03958755882069103,0.17669464777770372,-0.31934273661452756,0.12412443134582608,-0.18060188448566722,0.13113621027855027,0.007906533949455048,0.021387055172169718,0.02456848543445272,0.05392633967473781,-0.16010431701961533,-0.13926968221233477,0.013986512858663753,0.011205574726445205,-0.07369451391285031,0.028272590113174415,-0.16582180420694376,-0.2588577883232444,0.01068951715434718,-0.12656898820020862,-0.049617301151029004],\"type\":\"scatter3d\"},{\"hovertemplate\":\"color=trending<br>PC 1=%{x}<br>PC 2=%{y}<br>PC 3=%{z}<extra></extra>\",\"legendgroup\":\"trending\",\"marker\":{\"color\":\"#00cc96\",\"symbol\":\"circle\"},\"mode\":\"markers\",\"name\":\"trending\",\"scene\":\"scene\",\"showlegend\":true,\"x\":[-0.08550490391937846,-0.16069873841331925,-0.06510675783424055,-0.04814462890504494,-0.09943557288673367,-0.11296808219731096,-0.060459948937718944,-0.08277328727206905,-0.0820058104013457,-0.09436596516511744,-0.01248732661870415,-0.13530638281411408,-0.09348257657193802,-0.18944585334589958,-0.13346141156308144,-0.09175590064801144,-0.09988997247483418,-0.15811707131459857,-0.1809640787247893,-0.13438953776149157,-0.05551812745061811,-0.10166719752126785,-0.09113544238106745,-0.10296378727138562,-0.09230479595545542,-0.05678847403982305],\"y\":[0.03701566166906337,0.1428681320219042,-0.05970701833601,0.026277601385367538,0.02013329968775527,0.0294808116258612,-0.014165966396282235,0.05339886168651722,-0.02333444008520932,-0.0006842198161980193,0.3167655211643034,0.07955146197764489,0.06955806359147808,0.06801147651358677,0.07955190336849674,0.10689052058391245,0.06594300726215387,0.13687794464190653,0.13322505358953343,0.07004279546024829,0.031726600747736755,0.05946440002841148,0.03231812027473474,0.007167921887000549,0.02884363977150363,0.0449672516642361],\"z\":[-0.14078211208870836,-0.06416777702294298,-0.07030294655196004,-0.017154041910486577,-0.043738700039276164,0.011463300760907572,0.012144897975001115,-0.030546240977443778,-0.055775506978560235,0.04875229010864319,-0.10852385042848496,-0.011619967277668865,0.04488780276826087,-0.059369778126200305,-0.06509242912868042,-0.10470816685454547,-0.011517508895166001,0.004850117305445444,-0.051044017482741,-0.08850731020494468,-0.0843165637604285,0.002407051285305122,-0.09058746761510938,0.04884983136025424,-0.09047128705953693,-0.11685071406514273],\"type\":\"scatter3d\"},{\"hovertemplate\":\"color=money<br>PC 1=%{x}<br>PC 2=%{y}<br>PC 3=%{z}<extra></extra>\",\"legendgroup\":\"money\",\"marker\":{\"color\":\"#ab63fa\",\"symbol\":\"circle\"},\"mode\":\"markers\",\"name\":\"money\",\"scene\":\"scene\",\"showlegend\":true,\"x\":[-0.02101573950648556,-0.07945854630056896,-0.09238013191657468,-0.053111542023066394,0.08020289945368861,-0.06118814707657865,-0.08927888086127084,-0.10423947447865602,-0.10741346464548546,-0.07400437575370954,-0.10343934502706241,-0.06027567761497174,-0.033812948984860314,-0.05678193031394479,-0.02505204748564591,-0.07476983465916047,-0.07659415570098926,-0.03700881190189795,-0.039735641680401286,-0.08263610814759627,0.046954511248013274,0.0006505050617471405,-0.041245233987368964,-0.007278396670547528,-0.06508315466665204,-0.00725461056342858,-0.02719314730186496,0.05973987989092781,-0.06674669271202793],\"y\":[-0.032498852560659465,0.007679725468886863,0.01823976188464063,-0.011895922006395842,0.042969303340313666,0.030005900287110823,0.04764416968577303,-0.045124995190194674,0.027305064368612072,-0.02614927919430085,0.0015350768930944302,0.04407617276027206,-0.06303665984749561,-0.0650326352662638,-0.07367615202174728,-0.0531997281006664,0.004335577119919258,-0.004927742368327323,-0.020510657861388937,0.010211977271181389,-0.09334701816722846,0.001027032852208413,-0.04967774628180768,-0.0017871472218388646,-0.00981808574444552,0.02574918113417247,-0.08542327786943095,0.10426744034315001,-0.003395633554941218],\"z\":[0.010020417502244748,0.018449230835984973,0.036169867652595215,-0.009157105875684675,0.2165407357711325,0.027403614006534474,-0.027443143831416314,-0.07151983594635807,0.035882556826952856,-0.010177430063995454,0.034539170139382444,0.04773380727708293,0.35623432321268406,-0.053148965730091084,-0.03068166132957259,-0.08344192280857891,-0.010528687500525425,0.1813706506383462,0.12621796565399174,0.06175561061659172,0.06259089474775138,0.16941368603472495,-0.07744165701230364,-0.06259468040146546,0.028989342230075963,0.025632592436344224,0.4144658883303687,0.01651565419106518,0.04651611251495619],\"type\":\"scatter3d\"},{\"hovertemplate\":\"color=netflix<br>PC 1=%{x}<br>PC 2=%{y}<br>PC 3=%{z}<extra></extra>\",\"legendgroup\":\"netflix\",\"marker\":{\"color\":\"#FFA15A\",\"symbol\":\"circle\"},\"mode\":\"markers\",\"name\":\"netflix\",\"scene\":\"scene\",\"showlegend\":true,\"x\":[-0.06015834972268309,0.003433990763980952,0.06516238906195938,-0.017026324142035316,-0.06444608664176796,-0.0663712919392217,-0.13990469290710417,0.036071743376980686,-0.05625890533217918,-0.023748674592726716,-0.040256161930768995,-0.13931338164942564,-0.053458186550629255,0.14717418121309425,-0.04043921151775328,0.0443237223230412,-0.052793938351015,-0.06312138636441844,0.01651182843126747,-0.09166769471560024,0.06601247423531802,0.026066095257158087,0.05847554868831281,-0.12047254861589572,-0.11821344665415315,-0.08755150367548153,0.061752865547266306,-0.031127767050324228,-0.030405692806592322],\"y\":[0.03446459716058081,-0.05320287580563843,-0.009035374268420993,-0.08894241766173791,0.006972993460156755,-0.0010705878184030908,0.10807544815632882,-0.04681472576846182,0.0017439534260990342,-0.023795868824863585,-0.038196621294087026,0.0950281460118503,-0.027571427602391656,-0.003492148613006558,0.012925452816031465,-0.019604506083000985,0.002530981173136906,0.012472905940230391,0.30297220332338615,-0.00547067094598277,-0.07260082648956429,0.07013735543925637,0.014972921111864925,0.11441148851154372,0.00834478039418557,0.007707902873119458,-0.008496835918116138,0.022965427480140427,-0.05537426179639288],\"z\":[0.01826325386792964,0.11869126698081912,0.006410054978173344,0.09641432438964921,-0.016726341513125246,-0.019512285228510978,0.13897861118129168,-0.028936807443998444,0.002680219945243023,0.029498698838218953,-0.053913977176895375,0.1546230318728741,0.01968815070549194,0.19503150966041582,0.10039542052979601,0.18855561296582388,0.027356560826643405,0.06844984282781318,-0.018120747800918238,-0.050791119345012355,0.11875531586064043,0.04003310788761651,-0.015126458709089782,0.16361792810649267,-0.06580345700752,-0.061287160741687996,-0.010305239891934583,-0.005576674782232184,-0.08554156751255135],\"type\":\"scatter3d\"}],                        {\"template\":{\"data\":{\"bar\":[{\"error_x\":{\"color\":\"#2a3f5f\"},\"error_y\":{\"color\":\"#2a3f5f\"},\"marker\":{\"line\":{\"color\":\"#E5ECF6\",\"width\":0.5},\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"bar\"}],\"barpolar\":[{\"marker\":{\"line\":{\"color\":\"#E5ECF6\",\"width\":0.5},\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"barpolar\"}],\"carpet\":[{\"aaxis\":{\"endlinecolor\":\"#2a3f5f\",\"gridcolor\":\"white\",\"linecolor\":\"white\",\"minorgridcolor\":\"white\",\"startlinecolor\":\"#2a3f5f\"},\"baxis\":{\"endlinecolor\":\"#2a3f5f\",\"gridcolor\":\"white\",\"linecolor\":\"white\",\"minorgridcolor\":\"white\",\"startlinecolor\":\"#2a3f5f\"},\"type\":\"carpet\"}],\"choropleth\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"choropleth\"}],\"contour\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"contour\"}],\"contourcarpet\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"contourcarpet\"}],\"heatmap\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"heatmap\"}],\"heatmapgl\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"heatmapgl\"}],\"histogram\":[{\"marker\":{\"pattern\":{\"fillmode\":\"overlay\",\"size\":10,\"solidity\":0.2}},\"type\":\"histogram\"}],\"histogram2d\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"histogram2d\"}],\"histogram2dcontour\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"histogram2dcontour\"}],\"mesh3d\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"type\":\"mesh3d\"}],\"parcoords\":[{\"line\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"parcoords\"}],\"pie\":[{\"automargin\":true,\"type\":\"pie\"}],\"scatter\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatter\"}],\"scatter3d\":[{\"line\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatter3d\"}],\"scattercarpet\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattercarpet\"}],\"scattergeo\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattergeo\"}],\"scattergl\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattergl\"}],\"scattermapbox\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scattermapbox\"}],\"scatterpolar\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterpolar\"}],\"scatterpolargl\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterpolargl\"}],\"scatterternary\":[{\"marker\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"type\":\"scatterternary\"}],\"surface\":[{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"},\"colorscale\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"type\":\"surface\"}],\"table\":[{\"cells\":{\"fill\":{\"color\":\"#EBF0F8\"},\"line\":{\"color\":\"white\"}},\"header\":{\"fill\":{\"color\":\"#C8D4E3\"},\"line\":{\"color\":\"white\"}},\"type\":\"table\"}]},\"layout\":{\"annotationdefaults\":{\"arrowcolor\":\"#2a3f5f\",\"arrowhead\":0,\"arrowwidth\":1},\"autotypenumbers\":\"strict\",\"coloraxis\":{\"colorbar\":{\"outlinewidth\":0,\"ticks\":\"\"}},\"colorscale\":{\"diverging\":[[0,\"#8e0152\"],[0.1,\"#c51b7d\"],[0.2,\"#de77ae\"],[0.3,\"#f1b6da\"],[0.4,\"#fde0ef\"],[0.5,\"#f7f7f7\"],[0.6,\"#e6f5d0\"],[0.7,\"#b8e186\"],[0.8,\"#7fbc41\"],[0.9,\"#4d9221\"],[1,\"#276419\"]],\"sequential\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]],\"sequentialminus\":[[0.0,\"#0d0887\"],[0.1111111111111111,\"#46039f\"],[0.2222222222222222,\"#7201a8\"],[0.3333333333333333,\"#9c179e\"],[0.4444444444444444,\"#bd3786\"],[0.5555555555555556,\"#d8576b\"],[0.6666666666666666,\"#ed7953\"],[0.7777777777777778,\"#fb9f3a\"],[0.8888888888888888,\"#fdca26\"],[1.0,\"#f0f921\"]]},\"colorway\":[\"#636efa\",\"#EF553B\",\"#00cc96\",\"#ab63fa\",\"#FFA15A\",\"#19d3f3\",\"#FF6692\",\"#B6E880\",\"#FF97FF\",\"#FECB52\"],\"font\":{\"color\":\"#2a3f5f\"},\"geo\":{\"bgcolor\":\"white\",\"lakecolor\":\"white\",\"landcolor\":\"#E5ECF6\",\"showlakes\":true,\"showland\":true,\"subunitcolor\":\"white\"},\"hoverlabel\":{\"align\":\"left\"},\"hovermode\":\"closest\",\"mapbox\":{\"style\":\"light\"},\"paper_bgcolor\":\"white\",\"plot_bgcolor\":\"#E5ECF6\",\"polar\":{\"angularaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"bgcolor\":\"#E5ECF6\",\"radialaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"}},\"scene\":{\"xaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"},\"yaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"},\"zaxis\":{\"backgroundcolor\":\"#E5ECF6\",\"gridcolor\":\"white\",\"gridwidth\":2,\"linecolor\":\"white\",\"showbackground\":true,\"ticks\":\"\",\"zerolinecolor\":\"white\"}},\"shapedefaults\":{\"line\":{\"color\":\"#2a3f5f\"}},\"ternary\":{\"aaxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"baxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"},\"bgcolor\":\"#E5ECF6\",\"caxis\":{\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\"}},\"title\":{\"x\":0.05},\"xaxis\":{\"automargin\":true,\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\",\"title\":{\"standoff\":15},\"zerolinecolor\":\"white\",\"zerolinewidth\":2},\"yaxis\":{\"automargin\":true,\"gridcolor\":\"white\",\"linecolor\":\"white\",\"ticks\":\"\",\"title\":{\"standoff\":15},\"zerolinecolor\":\"white\",\"zerolinewidth\":2}}},\"scene\":{\"domain\":{\"x\":[0.0,1.0],\"y\":[0.0,1.0]},\"xaxis\":{\"title\":{\"text\":\"PC 1\"}},\"yaxis\":{\"title\":{\"text\":\"PC 2\"}},\"zaxis\":{\"title\":{\"text\":\"PC 3\"}}},\"legend\":{\"title\":{\"text\":\"color\"},\"tracegroupgap\":0},\"title\":{\"text\":\"Total explained variance: 3.70%\"}},                        {\"responsive\": true}                    ).then(function(){\n",
       "                            \n",
       "var gd = document.getElementById('75fd6a97-d7d0-47cb-ba2e-434d4d74fd03');\n",
       "var x = new MutationObserver(function (mutations, observer) {{\n",
       "        var display = window.getComputedStyle(gd).display;\n",
       "        if (!display || display === 'none') {{\n",
       "            console.log([gd, 'removed!']);\n",
       "            Plotly.purge(gd);\n",
       "            observer.disconnect();\n",
       "        }}\n",
       "}});\n",
       "\n",
       "// Listen for the removal of the full notebook cells\n",
       "var notebookContainer = gd.closest('#notebook-container');\n",
       "if (notebookContainer) {{\n",
       "    x.observe(notebookContainer, {childList: true});\n",
       "}}\n",
       "\n",
       "// Listen for the clearing of the current output cell\n",
       "var outputEl = gd.closest('.output');\n",
       "if (outputEl) {{\n",
       "    x.observe(outputEl, {childList: true});\n",
       "}}\n",
       "\n",
       "                        })                };                            </script>        </div>\n",
       "</body>\n",
       "</html>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pca = PCA(n_components=3)\n",
    "components = pca.fit_transform(dtm_df)\n",
    "total_var = pca.explained_variance_ratio_.sum() * 100\n",
    "px.scatter_3d(\n",
    "    components, x=0, y=1, z=2, color=sample_df['forum_en'],\n",
    "    title=f'Total explained variance: {total_var:.2f}%',\n",
    "    labels={'0': 'PC 1', '1': 'PC 2', '2': 'PC 3'}\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {
    "id": "LL_oz8AXt37J"
   },
   "outputs": [],
   "source": [
    "from gensim import similarities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "id": "oeu2TKmet5Cj"
   },
   "outputs": [],
   "source": [
    "def search_by_tfidf(tokenized_text: list) -> tuple:\n",
    "  vec_bow = dictionary.doc2bow(tokenized_text)\n",
    "  vec_tfidf = tfidf[vec_bow]\n",
    "  index = similarities.MatrixSimilarity(tfidf[corpus])  \n",
    "  sims = index[vec_tfidf]  \n",
    "  sims = sorted(enumerate(sims), key=lambda item: item[1], reverse=True)\n",
    "  indexes = [tup[0] for tup in sims]\n",
    "  scores = [tup[1] for tup in sims]\n",
    "  return (indexes, scores)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vrJJ2eJ-3UVS"
   },
   "source": [
    "- You can change the value of `docid` to any number between 0 and 149."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 641
    },
    "id": "KzD8H43quNUU",
    "outputId": "bd069c9c-a7cb-4bb7-9e69-3e70fd23f16e"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-9d5d1468-e783-488c-9355-85dc94b9da30\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>excerpt</th>\n",
       "      <th>forum_en</th>\n",
       "      <th>forum_zh</th>\n",
       "      <th>tfidf_sim</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>73</th>\n",
       "      <td>一個是等烏克蘭首都被攻下，一個是等布丁死亡，有俄羅斯自家富豪出手了，BTW俄羅斯自家的駭客也...</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>57</th>\n",
       "      <td>現在最多支持俄羅斯發起戰爭的理由就是這個了，假設今天俄羅斯打贏了，烏克蘭歸屬於俄羅斯，結果北...</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>0.253894</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>74</th>\n",
       "      <td>全台大停電我只想只有一個原因！，美國看到俄羅斯就攻不下烏克蘭，想要在俄羅斯自顧不暇之餘誘使中...</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>0.119545</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>68</th>\n",
       "      <td>誒，烏俄戰爭都打這麼久了 ，世界各國只是物資上的幫忙，是想要看烏克蘭全體軍民英勇殉國嗎？</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>0.083784</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>70</th>\n",
       "      <td>想必有操作金融市場的朋友，不管是股市或幣圈，今早都來了一個震撼彈。台股指數跳空開低4百點，幣...</td>\n",
       "      <td>trending</td>\n",
       "      <td>時事</td>\n",
       "      <td>0.069376</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>128</th>\n",
       "      <td>《少年法院》的原聲帶配樂由音樂導演金泰成監製，電影有「1987：黎明到來的那一天」「黑祭司」...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>130</th>\n",
       "      <td>第一次看到城市名稱這麼有感覺的 好好笑，真的好捨不得結束哦嗚嗚嗚，聽說韓國要翻拍了，期待期待！</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136</th>\n",
       "      <td>《二十五，二十一》第五首原聲帶歌曲由男團SEVENTEEN的DK來擔任，在當DK要演唱原聲帶...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>137</th>\n",
       "      <td>我想問有沒有類似荒謬分局那種一集時間不長，但是很耐看，很多季的影集嗎（；´༎ຶД༎ຶ`），最...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>138</th>\n",
       "      <td>借板發一下，感謝。1這部老實講比正傳悶，悶很多，有點小失望，看前三集除了格陵蘭人渡海那段，其...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>140 rows × 4 columns</p>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-9d5d1468-e783-488c-9355-85dc94b9da30')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-9d5d1468-e783-488c-9355-85dc94b9da30 button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-9d5d1468-e783-488c-9355-85dc94b9da30');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                               excerpt  forum_en forum_zh  \\\n",
       "73   一個是等烏克蘭首都被攻下，一個是等布丁死亡，有俄羅斯自家富豪出手了，BTW俄羅斯自家的駭客也...  trending       時事   \n",
       "57   現在最多支持俄羅斯發起戰爭的理由就是這個了，假設今天俄羅斯打贏了，烏克蘭歸屬於俄羅斯，結果北...  trending       時事   \n",
       "74   全台大停電我只想只有一個原因！，美國看到俄羅斯就攻不下烏克蘭，想要在俄羅斯自顧不暇之餘誘使中...  trending       時事   \n",
       "68        誒，烏俄戰爭都打這麼久了 ，世界各國只是物資上的幫忙，是想要看烏克蘭全體軍民英勇殉國嗎？  trending       時事   \n",
       "70   想必有操作金融市場的朋友，不管是股市或幣圈，今早都來了一個震撼彈。台股指數跳空開低4百點，幣...  trending       時事   \n",
       "..                                                 ...       ...      ...   \n",
       "128  《少年法院》的原聲帶配樂由音樂導演金泰成監製，電影有「1987：黎明到來的那一天」「黑祭司」...   netflix  Netflix   \n",
       "130    第一次看到城市名稱這麼有感覺的 好好笑，真的好捨不得結束哦嗚嗚嗚，聽說韓國要翻拍了，期待期待！   netflix  Netflix   \n",
       "136  《二十五，二十一》第五首原聲帶歌曲由男團SEVENTEEN的DK來擔任，在當DK要演唱原聲帶...   netflix  Netflix   \n",
       "137  我想問有沒有類似荒謬分局那種一集時間不長，但是很耐看，很多季的影集嗎（；´༎ຶД༎ຶ`），最...   netflix  Netflix   \n",
       "138  借板發一下，感謝。1這部老實講比正傳悶，悶很多，有點小失望，看前三集除了格陵蘭人渡海那段，其...   netflix  Netflix   \n",
       "\n",
       "     tfidf_sim  \n",
       "73    1.000000  \n",
       "57    0.253894  \n",
       "74    0.119545  \n",
       "68    0.083784  \n",
       "70    0.069376  \n",
       "..         ...  \n",
       "128   0.000000  \n",
       "130   0.000000  \n",
       "136   0.000000  \n",
       "137   0.000000  \n",
       "138   0.000000  \n",
       "\n",
       "[140 rows x 4 columns]"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "docid = 73\n",
    "tokens_in_raw_text = sample_df.loc[docid,'toks']\n",
    "indexes, scores = search_by_tfidf(tokens_in_raw_text)\n",
    "results_df = sample_df.loc[indexes][['excerpt', 'forum_en', 'forum_zh']]\n",
    "results_df['tfidf_sim'] = scores\n",
    "results_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Pu-zI1kx5kB1",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Distributional representation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "v-_6NTbOm4m3"
   },
   "source": [
    "> Distributional hypothesis: Linguistic items with similar distributions have similar meanings. [(Wikipedia)](https://en.wikipedia.org/wiki/Distributional_semantics)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9VrjxDV1Ounm"
   },
   "source": [
    "![](https://slideplayer.com/slide/12147948/71/images/10/John+Rupert+Firth+You+shall+know+a+word+by+the+company+it+keeps.jpg)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8xImkr2Rpzoa"
   },
   "source": [
    "### word2vec"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vrDXEni7urxs"
   },
   "source": [
    "- Input: number of all unique words\n",
    "- Output: dimensions of a word vector\n",
    "![](https://miro.medium.com/max/1000/1*1O5gLhOg25HviI8bwZxV4g.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "T452ZOBJTyy-"
   },
   "source": [
    "- Window size: 5\n",
    "- Training mission: predict the targt word from context words\n",
    "- Target word: the word in blue\n",
    "- Context words: the words in red\n",
    "\n",
    "![](https://twice22.github.io/images/word2vec/cbow-example.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "nyntvYrEAINY"
   },
   "source": [
    "- A token is represented by a 300-dimension dense vector."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zIAjhmYl4_US",
    "outputId": "cb5c2deb-50f3-48c8-acd7-ee40f1bbe3db"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 2.2328  , -1.1713  , -3.3528  , -1.1691  , -0.26724 ,  4.4476  ,\n",
       "       -0.66089 ,  2.6248  , -1.5367  , -2.8449  , -4.0233  ,  1.5727  ,\n",
       "        1.978   ,  2.7964  ,  1.003   ,  0.29978 ,  0.056525,  3.7048  ,\n",
       "        2.0446  ,  2.2452  , -5.7184  ,  0.77814 , -1.8383  , -0.017231,\n",
       "       -1.91    , -6.4355  , -4.6737  , -0.13519 ,  0.66087 , -1.6718  ,\n",
       "        3.5934  ,  2.3382  , -4.5406  ,  1.6124  , -2.2361  , -6.0387  ,\n",
       "       -3.4078  ,  1.1304  ,  0.80933 ,  1.9734  ,  2.3314  , -0.9882  ,\n",
       "       -1.1947  ,  2.2628  , -1.3687  , -6.4278  ,  0.15906 ,  0.047335,\n",
       "       -2.8157  , -1.6407  ,  2.4385  , -0.84336 ,  3.081   ,  5.9188  ,\n",
       "       -1.3019  ,  1.2971  ,  7.2325  ,  2.9722  , -0.45552 ,  1.5148  ,\n",
       "       -1.1193  ,  3.8739  ,  1.482   , -2.4657  ,  1.4627  , -3.562   ,\n",
       "       -2.1737  , -1.4306  ,  3.4363  , -1.2796  , -1.4106  ,  2.2146  ,\n",
       "        2.9325  , -2.5172  ,  2.7192  , -0.84556 , -2.5362  ,  2.2079  ,\n",
       "       -3.2217  , -2.2081  ,  4.6204  ,  0.98445 , -0.53713 ,  1.5325  ,\n",
       "        0.60471 ,  1.774   ,  2.5593  ,  4.8882  , -2.9012  ,  0.56529 ,\n",
       "       -0.16369 , -2.6124  , -0.89601 ,  3.6279  , -3.0837  ,  7.4534  ,\n",
       "        5.5267  ,  0.30507 , -0.21904 ,  2.1489  , -3.5027  , -2.8241  ,\n",
       "        1.7284  , -1.7447  ,  0.05638 , -7.6138  , -3.8239  , -0.44218 ,\n",
       "        2.7571  ,  2.5495  , -3.6081  , -0.30223 ,  2.2124  , -0.82708 ,\n",
       "       -1.4102  , -0.81168 , -1.8045  ,  1.6797  , -0.43188 , -2.769   ,\n",
       "       -4.9266  ,  1.0018  ,  0.022032,  1.6223  , -0.097899,  1.541   ,\n",
       "       -4.1115  , -4.2305  , -2.0175  , -2.7113  ,  1.8462  ,  1.115   ,\n",
       "       -2.4203  ,  7.1492  , -1.3467  ,  5.1815  , -1.6418  , -0.90401 ,\n",
       "       -0.030428,  2.4756  , -3.3246  ,  1.2715  , -3.9147  , -2.4505  ,\n",
       "       -1.9256  ,  0.61014 ,  1.52    ,  0.20039 ,  6.3868  , -3.6932  ,\n",
       "       -0.25369 ,  1.993   ,  1.2806  , -1.4946  , -3.8367  ,  0.44791 ,\n",
       "        2.2502  ,  2.687   , -1.9332  ,  1.6963  ,  3.1247  , -4.6094  ,\n",
       "       -0.90111 ,  4.4618  , -5.5256  , -0.044889, -1.496   ,  2.8646  ,\n",
       "        1.9181  ,  0.82574 ,  0.024646, -4.0564  ,  1.1392  , -1.8365  ,\n",
       "       -0.38447 ,  0.063265,  1.2402  , -0.45659 , -1.7064  , -3.4588  ,\n",
       "       -0.23503 , -3.1285  , -2.8034  ,  2.1534  , -1.2622  , -4.0139  ,\n",
       "       -3.3527  ,  3.9156  ,  4.1822  , -1.1544  , -1.8433  , -3.4078  ,\n",
       "        0.43852 , -1.3131  , -0.92715 ,  4.6758  ,  1.8561  ,  3.4259  ,\n",
       "        1.2407  , -0.10015 , -0.4736  , -1.1884  ,  2.5849  , -1.7642  ,\n",
       "       -5.4619  ,  2.8648  ,  5.4742  , -1.9069  ,  0.87298 , -3.3007  ,\n",
       "        0.64926 ,  1.2753  ,  0.57441 ,  5.3602  ,  3.4332  , -3.4726  ,\n",
       "        0.41052 , -4.5747  ,  1.3222  ,  0.64865 , -1.7875  ,  0.68921 ,\n",
       "       -0.16166 ,  2.2868  ,  0.91401 ,  1.388   ,  7.0462  , -1.5426  ,\n",
       "        0.47719 , -3.237   , -1.0982  ,  4.7626  ,  3.4092  , -0.2124  ,\n",
       "       -0.4526  ,  4.4247  , -0.83311 , -0.99074 ,  2.1257  , -0.91728 ,\n",
       "        2.3176  ,  0.91452 , -1.436   ,  2.2201  ,  0.90931 , -0.46145 ,\n",
       "       -1.7671  , -1.594   ,  1.4875  , -3.7585  , -3.826   ,  4.4597  ,\n",
       "       -8.0136  , -1.7391  , -4.4454  ,  1.2797  ,  1.2241  , -0.85194 ,\n",
       "        0.42698 , -0.9868  , -5.0362  , -6.1074  ,  3.8487  ,  3.2342  ,\n",
       "        1.6124  , -1.2368  , -1.0743  , -1.1119  ,  2.7379  ,  4.4379  ,\n",
       "        3.8747  ,  0.83282 , -4.6553  , -1.303   , -0.27529 , -1.2466  ,\n",
       "       -1.5356  , -4.7198  , -5.2186  ,  1.2375  ,  1.3111  , -0.26792 ,\n",
       "        0.79845 , -0.25837 ,  2.2433  , -2.5042  ,  2.5327  ,  1.6947  ,\n",
       "       -3.1678  , -0.55927 , -1.17    , -1.2356  ,  0.3977  ,  1.8565  ,\n",
       "       -1.4511  ,  2.0813  ,  1.5979  , -0.50737 ,  2.5565  ,  0.63113 ],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "doc = nlp(\"教授\")\n",
    "tok = doc[0]\n",
    "tok.vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Ki7NEt5G5J8X",
    "outputId": "8737fef6-88fb-4b33-cae2-e02da2132a56"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(300,)"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tok.vector.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9sJXminV-TJq"
   },
   "source": [
    "- You can try to replace the values of `word_1`, `word_2`, and `word_3` to other Chinese words."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Fzf2Dyi46Ciz",
    "outputId": "d02cbed4-872b-45b6-a1ad-f614f7ff5419"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Distance btn Word 1 and 2: 0.27085748314857483\n",
      "Distance btn Word 1 and 3: 0.8141297101974487\n"
     ]
    }
   ],
   "source": [
    "word_1 = nlp.vocab[\"高興\"]\n",
    "word_2 = nlp.vocab[\"高雄\"]\n",
    "word_3 = nlp.vocab[\"開心\"]\n",
    "word_1_word_2 = word_1.similarity(word_2)\n",
    "word_1_word_3 = word_1.similarity(word_3)\n",
    "print(f\"Distance btn Word 1 and 2: {word_1_word_2}\")\n",
    "print(f\"Distance btn Word 1 and 3: {word_1_word_3}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "oOVU5VCCq1f1"
   },
   "source": [
    "- Cosine similarity \n",
    "\n",
    "![](https://datascience-enthusiast.com/figures/cosine_sim.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KS14jgQ190Mn"
   },
   "source": [
    "- Formula for calculating cosine similarity between two vectors\n",
    "\n",
    "![](https://miro.medium.com/max/1400/1*LfW66-WsYkFqWc4XYJbEJg.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_2tIZ_U-Atex"
   },
   "source": [
    "- You can replace the numbers `79`, `80`, and `139` with any number between 0 and 139."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7-rEkZAo_SKx",
    "outputId": "8220ae9c-3fe5-4fbf-abf1-29964c8482f6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Text 1 > 一個是等烏克蘭首都被攻下，一個是等布丁死亡，有俄羅斯自家富豪出手了，BTW俄羅斯自家的駭客也在幫忙打布丁\n",
      "Text 2 > 全台大停電我只想只有一個原因！，美國看到俄羅斯就攻不下烏克蘭，想要在俄羅斯自顧不暇之餘誘使中共主動出擊，龐佩奧才和蔡總統裡應外合演了停電的一檔大戲！，雖然習包子害怕自己落的跟普丁一樣的處境，在台灣大停\n",
      "Text 3 > #重雷，因為版上都是讚美的聲音…，其實幾位主角性格都很討喜，只是太多人大大大推了 個人看了反而有點小失望，第一集開頭用2020疫情和1997金融海嘯來提到，「時代影響了很多人」也覺得很有共鳴，因為我們\n"
     ]
    }
   ],
   "source": [
    "text_1 = sample_df.loc[73, \"excerpt\"]\n",
    "text_2 = sample_df.loc[74, \"excerpt\"]\n",
    "text_3 = sample_df.loc[139, \"excerpt\"]\n",
    "print(f\"Text 1 > {text_1}\")\n",
    "print(f\"Text 2 > {text_2}\")\n",
    "print(f\"Text 3 > {text_3}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "tPlwok0w-rMy",
    "outputId": "48f36583-2425-4542-fdd2-3640897dcff9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Distance btn Text 1 and 2: 0.9008373137883697\n",
      "Distance btn Text 1 and 3: 0.8674700824214483\n"
     ]
    }
   ],
   "source": [
    "doc_1 = nlp(text_1)\n",
    "doc_2 = nlp(text_2)\n",
    "doc_3 = nlp(text_3)\n",
    "doc_1_doc_2 = doc_1.similarity(doc_2)\n",
    "doc_1_doc_3 = doc_1.similarity(doc_3)\n",
    "print(f\"Distance btn Text 1 and 2: {doc_1_doc_2}\")\n",
    "print(f\"Distance btn Text 1 and 3: {doc_1_doc_3}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DYZKLawWCEvg"
   },
   "source": [
    "- If you're trying to be creative, feel free to replace the sentences for `doc01`, `doc02`, and `doc03` with any Chinese texts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "frKEFxotBsPE",
    "outputId": "532d5879-bf86-4fdc-ad2c-4a109231d052"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Distance btn Text 1 and 2: 0.5731685167538273\n",
      "Distance btn Text 1 and 3: 0.4037113663445769\n"
     ]
    }
   ],
   "source": [
    "doc01 = nlp(\"防Omicron威脅 一至三類今起可打第3劑\")\n",
    "doc02 = nlp(\"第3劑開打…賣場設站隨到隨打 首波4.8萬人符資格\")\n",
    "doc03 = nlp(\"西半部低溫特報 全台平地最冷在基隆僅8度\")\n",
    "sim01 = doc01.similarity(doc02)\n",
    "sim02 = doc01.similarity(doc03)\n",
    "print(f\"Distance btn Text 1 and 2: {sim01}\")\n",
    "print(f\"Distance btn Text 1 and 3: {sim02}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "cbZ5DtHe5oyt",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Distributed representation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dcCvPiO7IDtB"
   },
   "source": [
    "### Transformer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "76piHJgqIF0Z"
   },
   "source": [
    "- Transformer is a deep learning architecture for transforming one sequence to another. \n",
    "- It's the foundation of many state-of-the-art models, including BERT and GPT.\n",
    "\n",
    "![](https://factored.ai/wp-content/uploads/2021/09/image4.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Ph-ej5e1JHJA"
   },
   "source": [
    "- Disadvantages of distributional representation like word2vec\n",
    "    - A given token is represented by the same vector irrespective of its context.\n",
    "    ![](https://cdn.cultofmac.com/wp-content/uploads/2013/08/tumblr_li341fpH611qe0hneo1_1280.jpg)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WItMdwsKNy-z"
   },
   "source": [
    "- Disadvantages of distributional representation like word2vec\n",
    "    - A word out of the vocabulary scope has no vector.\n",
    "    ![](https://cdn1.vectorstock.com/i/1000x1000/22/85/404-error-page-not-found-with-computer-vector-26642285.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "quXLd07yLpLS"
   },
   "source": [
    "- Disadvantages of distributional representation like word2vec\n",
    "    - The relationship holding between words may go beyond the window size.\n",
    "    ![](https://pbs.twimg.com/media/EkK0r3TXgAIFZoM.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zhEQ_bOMM779"
   },
   "source": [
    "- Transformer solves all these problems and its killing feature is self-attention.\n",
    "![](https://jalammar.github.io/images/t/transformer_self-attention_visualization_3.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "STGiQDMlPDP6",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Transfer learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "mG8WT_W1P3ce"
   },
   "source": [
    "![](https://img.favpng.com/17/24/20/transfer-learning-machine-learning-teacher-deep-learning-png-favpng-zXhj8f0jTbRAiXtvvGdqJZvjc.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YKQ8vCfBv2Qm",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Intuition of transfer learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "A3Utnvjhh9gT"
   },
   "source": [
    "- Transfer learning taps into a skill previously learned when one tries to learn a new skill.\n",
    "\n",
    "![](http://www.blog.sagmart.com/wp-content/uploads/2014/07/Bicycle-VS-Motorcycle.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8qCKsDEnimyH"
   },
   "source": [
    "- In traditional machine learning, a model learns from scratch.\n",
    "- In transfer learning, a model reuses the knowledge that a previously trained model has learned.\n",
    "\n",
    "![](https://www.mdpi.com/electronics/electronics-10-01491/article_deploy/html/images/electronics-10-01491-g001.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Dc0D141Dv5un"
   },
   "source": [
    "- In deep learning, the knowledge gained in a trained model is mathematically weights of model parameters.\n",
    "\n",
    "![](https://miro.medium.com/max/1000/1*mA1sUreCxnl-65ljlaXEcA.jpeg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2WQmsVkltTfc"
   },
   "source": [
    "- In NLP, various source models are quite effective in transfering knowledge to target models, including word2vec and BERT.\n",
    "![](https://ruder.io/content/images/2019/08/pretraining_adaptation.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "pLmFS1EJnR3I"
   },
   "source": [
    "- With transfer learning, we can easily stand on the shoulders of giants.\n",
    "\n",
    "![](https://miro.medium.com/max/910/1*J9vRAiAxtOV3dJGxhS518A.jpeg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "htdkif3DpwPH",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Types of transfer learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qzTKcH1z1R6D"
   },
   "source": [
    "[(Ruder 2019)](https://ruder.io/thesis/neural_transfer_learning_for_nlp.pdf#page=64)\n",
    "![](https://ruder.io/content/images/2019/08/transfer_learning_taxonomy.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6q25ZSEhj-pl"
   },
   "source": [
    "- Domain adaptation\n",
    "    - Train on Wikipedia texts\n",
    "    - Predict on tweets  \n",
    "![](https://miro.medium.com/max/1400/1*OhB35eeMnGm99xhBA9A1vA.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5ZCCRRKHkYV_"
   },
   "source": [
    "- Cross-lingual learning\n",
    "    - Train on English\n",
    "    - Predict on Spanish\n",
    "![](https://www.researchgate.net/profile/Jeremy-Barnes-2/publication/309312650/figure/fig1/AS:669424235323406@1536614583578/The-process-of-cross-lingual-sentiment-classification-We-assume-that-the-opinion-units.ppm)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Ox-BkiX2rxBw"
   },
   "source": [
    "- Multi-task learning [(Liu et al. 2019)](https://arxiv.org/abs/1901.11504)\n",
    "\n",
    "![](https://www.researchgate.net/profile/Xiaodong-Liu-35/publication/332590404/figure/fig1/AS:750869628870659@1556032678647/Architecture-of-the-MT-DNN-model-for-representation-learning-Liu-et-al-2019-The.ppm)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "LXYIXFyDtPWT"
   },
   "source": [
    "- Sequential learning\n",
    "    - BERT is trained to \n",
    "        - predict a missing token in a random position\n",
    "        - predict whether a sentence follows another\n",
    "    - Use BERT's knowledge to train a new model for a new skill \n",
    "\n",
    "![](https://www.researchgate.net/profile/Farhad-Nooralahzadeh-2/publication/345655007/figure/fig5/AS:956170839810048@1604980301263/Sequential-transfer-learning-using-BERT-left-Pre-training-stage-using-transformer.ppm)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Cj8uYd4JqEjS"
   },
   "source": [
    "- spaCy's framework makes it easy to do sequential transfer learning.\n",
    "\n",
    "![](https://spacy.io/tok2vec-listener-8c4d53807708b270c07a085f4a2da75f.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "iPEOKDWxlfK8",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Hands-on walkthrough (HOW)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "L6M5IawXZNpK"
   },
   "source": [
    "![](https://www.1440.org/resourcefiles/blogsmallimages/blog-jun13-1920x576.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Te_DYPC39ZDQ",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 762
    },
    "id": "0gDdGEhkV3L_",
    "outputId": "2b1e5896-3d29-4439-d63a-0fb06a21e74d"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-01048a66-288a-44c9-9b00-23650e3f496f\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>excerpt</th>\n",
       "      <th>forum_en</th>\n",
       "      <th>forum_zh</th>\n",
       "      <th>toks</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>希望各位能花個20秒幫我填一下</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[希望, 位, 花, 個, 秒, 幫, 填]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[想, 找, 這, 套, 衣服, 🥲, 發現, 該, 用什麼, 關鍵字, 找, 圖是, 草屯...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[因為, 文會, 有點, 長, 先, 說, 結論, 網購, 過, 平台, 退貨, 麻煩, 家...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[來源, 覺得, 呱吉, 這, 襯衫, 好好, 類似, 嗎]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...</td>\n",
       "      <td>dressup</td>\n",
       "      <td>穿搭</td>\n",
       "      <td>[位, 因為, 這, 個, 證件, 夾, 臺灣, 買不到, 美國, 限量版, 貨, 這, 間...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>135</th>\n",
       "      <td>最初只是和平的人權運動，最後卻演變成一場血腥暴力的革命，烏克蘭人民用93天讓世界見證了他們的...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>[最初, 和平, 人權, 運動, 卻, 演變成, 一場, 血腥, 暴力, 革命, 烏克蘭, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>136</th>\n",
       "      <td>《二十五，二十一》第五首原聲帶歌曲由男團SEVENTEEN的DK來擔任，在當DK要演唱原聲帶...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>[首, 原, 聲帶, 歌曲, 男團, 來, 擔任, 當, 演唱, 原, 聲帶, 消息, 時,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>137</th>\n",
       "      <td>我想問有沒有類似荒謬分局那種一集時間不長，但是很耐看，很多季的影集嗎（；´༎ຶД༎ຶ`），最...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>[想問, 有沒有, 類似, 荒謬, 分局, 種, 一集, 時間, 不長, 耐看, 季, 影集...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>138</th>\n",
       "      <td>借板發一下，感謝。1這部老實講比正傳悶，悶很多，有點小失望，看前三集除了格陵蘭人渡海那段，其...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>[板發, 感謝, 這, 部, 老實, 講, 正傳, 悶, 悶, 有點, 失望, 前三集, 格...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>139</th>\n",
       "      <td>#重雷，因為版上都是讚美的聲音…，其實幾位主角性格都很討喜，只是太多人大大大推了 個人看了反...</td>\n",
       "      <td>netflix</td>\n",
       "      <td>Netflix</td>\n",
       "      <td>[重雷, 因為, 版上, 讚美, 聲音, 其實, 幾, 位, 主角, 性格, 討喜, 太多,...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>140 rows × 4 columns</p>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-01048a66-288a-44c9-9b00-23650e3f496f')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-01048a66-288a-44c9-9b00-23650e3f496f button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-01048a66-288a-44c9-9b00-23650e3f496f');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "                                               excerpt forum_en forum_zh  \\\n",
       "0                                      希望各位能花個20秒幫我填一下  dressup       穿搭   \n",
       "1              想找這套衣服🥲，但發現不知道該用什麼關鍵字找，（圖是草屯囝仔的校園演唱會截圖）  dressup       穿搭   \n",
       "2    因為文會有點長，先說結論是，50%是目前網購過的平台退貨最麻煩的一家，甚至我認為根本是刻意刁...  dressup       穿搭   \n",
       "3                           來源：覺得呱吉這襯衫好好看~~，或有人知道有類似的嗎  dressup       穿搭   \n",
       "4    各位，因為這個證件夾臺灣買不到，是美國outlet 的限量版貨，所以在以下的這間蝦皮上買，但...  dressup       穿搭   \n",
       "..                                                 ...      ...      ...   \n",
       "135  最初只是和平的人權運動，最後卻演變成一場血腥暴力的革命，烏克蘭人民用93天讓世界見證了他們的...  netflix  Netflix   \n",
       "136  《二十五，二十一》第五首原聲帶歌曲由男團SEVENTEEN的DK來擔任，在當DK要演唱原聲帶...  netflix  Netflix   \n",
       "137  我想問有沒有類似荒謬分局那種一集時間不長，但是很耐看，很多季的影集嗎（；´༎ຶД༎ຶ`），最...  netflix  Netflix   \n",
       "138  借板發一下，感謝。1這部老實講比正傳悶，悶很多，有點小失望，看前三集除了格陵蘭人渡海那段，其...  netflix  Netflix   \n",
       "139  #重雷，因為版上都是讚美的聲音…，其實幾位主角性格都很討喜，只是太多人大大大推了 個人看了反...  netflix  Netflix   \n",
       "\n",
       "                                                  toks  \n",
       "0                               [希望, 位, 花, 個, 秒, 幫, 填]  \n",
       "1    [想, 找, 這, 套, 衣服, 🥲, 發現, 該, 用什麼, 關鍵字, 找, 圖是, 草屯...  \n",
       "2    [因為, 文會, 有點, 長, 先, 說, 結論, 網購, 過, 平台, 退貨, 麻煩, 家...  \n",
       "3                       [來源, 覺得, 呱吉, 這, 襯衫, 好好, 類似, 嗎]  \n",
       "4    [位, 因為, 這, 個, 證件, 夾, 臺灣, 買不到, 美國, 限量版, 貨, 這, 間...  \n",
       "..                                                 ...  \n",
       "135  [最初, 和平, 人權, 運動, 卻, 演變成, 一場, 血腥, 暴力, 革命, 烏克蘭, ...  \n",
       "136  [首, 原, 聲帶, 歌曲, 男團, 來, 擔任, 當, 演唱, 原, 聲帶, 消息, 時,...  \n",
       "137  [想問, 有沒有, 類似, 荒謬, 分局, 種, 一集, 時間, 不長, 耐看, 季, 影集...  \n",
       "138  [板發, 感謝, 這, 部, 老實, 講, 正傳, 悶, 悶, 有點, 失望, 前三集, 格...  \n",
       "139  [重雷, 因為, 版上, 讚美, 聲音, 其實, 幾, 位, 主角, 性格, 討喜, 太多,...  \n",
       "\n",
       "[140 rows x 4 columns]"
      ]
     },
     "execution_count": 85,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols = [\"excerpt\", \"forum_en\", \"forum_zh\", \"toks\"]\n",
    "sample_df[cols]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-HD7iSDvjK2f"
   },
   "source": [
    "## Two text classification models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hYZ-wrvF42u6"
   },
   "source": [
    "- Classification is a very common task. \n",
    "\n",
    "- Many problems can be cast as a classification task, including\n",
    "    - Sentiment analysis \n",
    "    - Part-of-speech tagging\n",
    "    - Dependency parsing\n",
    "    - Named entity recognition\n",
    "    - Intent recognition for chatbots \n",
    "\n",
    "- We'll train two classification models that take a text as input and predict a label for it.\n",
    "    - TF-IDF: no transfer learning\n",
    "    - Transformer: leveraging transfer learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sja1AIgm104E",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### TF-IDF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {
    "id": "9dtULrFQ92Cd"
   },
   "outputs": [],
   "source": [
    "from sklearn import feature_extraction, model_selection, pipeline\n",
    "from sklearn.svm import SVC \n",
    "from sklearn import metrics\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "id": "5TchZgMy93rN"
   },
   "outputs": [],
   "source": [
    "train_df, test_df = model_selection.train_test_split(sample_df, test_size=0.3, random_state=100)\n",
    "train_y = train_df[\"forum_en\"].values\n",
    "test_y = test_df[\"forum_en\"].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "RGdiIoHbXLkZ",
    "outputId": "d7e97974-7190-420a-8fc2-57922d99bb9f"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "forum_en\n",
       "dressup     24\n",
       "makeup      17\n",
       "money       17\n",
       "netflix     17\n",
       "trending    23\n",
       "dtype: int64"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.groupby(\"forum_en\").size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "G_r5G97MXNx5",
    "outputId": "6884315a-a448-4ec0-a168-035c0b91afec"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "forum_en\n",
       "dressup      4\n",
       "makeup      11\n",
       "money       12\n",
       "netflix     12\n",
       "trending     3\n",
       "dtype: int64"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_df.groupby(\"forum_en\").size()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hy8fj6LFaOlJ"
   },
   "source": [
    "- Let's use Support Vector Machine (SVC) for the classification algorithm.\n",
    "- SVC projects data in the input space to a feature space in higher dimensioons.\n",
    "\n",
    "![](https://qph.fs.quoracdn.net/main-qimg-cd6cde306c8273b2af183f57f25c259d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {
    "id": "Wy_cDfEXbR_a"
   },
   "outputs": [],
   "source": [
    "vectorizer = feature_extraction.text.TfidfVectorizer(max_features=2000, ngram_range=(1,2))\n",
    "corpus = train_df[\"toks\"].apply(lambda x: \" \".join(x))\n",
    "vectorizer.fit(corpus)\n",
    "train_X = vectorizer.transform(corpus)\n",
    "svc_clf = SVC(C=1, gamma=\"auto\", kernel='linear',probability=False)\n",
    "model = pipeline.Pipeline([(\"vectorizer\", vectorizer),  \n",
    "                           (\"svm\", svc_clf)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "5tyKdUKVas-8",
    "outputId": "05bb5f37-797b-49cc-8f37-e0af6e39282e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "It took 0.005840778350830078 secs to train the model.\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "model[\"svm\"].fit(train_X, train_y)\n",
    "end = time.time()\n",
    "print(f\"It took {end-start} secs to train the model.\") "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "kp1e-cXGcCoM",
    "outputId": "8f3dfebb-b465-4e62-ed36-a16030b067af"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "It took 0.005638599395751953 secs to make predictions on the testing data.\n"
     ]
    }
   ],
   "source": [
    "test_X = test_df[\"toks\"].apply(lambda x: \" \".join(x))\n",
    "start = time.time()\n",
    "predictions = model.predict(test_X)\n",
    "end = time.time()\n",
    "print(f\"It took {end-start} secs to make predictions on the testing data.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "bnXw3uEvfSo9",
    "outputId": "e2111b1b-82f9-48cc-a0cd-a9d6f5d60284"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['dressup', 'dressup', 'netflix', 'money', 'trending', 'dressup',\n",
       "       'makeup', 'dressup', 'dressup', 'makeup', 'dressup', 'trending',\n",
       "       'makeup', 'dressup', 'makeup', 'dressup', 'money', 'netflix',\n",
       "       'makeup', 'dressup', 'trending', 'money', 'netflix', 'netflix',\n",
       "       'dressup', 'dressup', 'money', 'dressup', 'dressup', 'dressup',\n",
       "       'money', 'trending', 'dressup', 'dressup', 'money', 'trending',\n",
       "       'makeup', 'dressup', 'dressup', 'trending', 'trending', 'trending'],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 169,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 300
    },
    "id": "5oZYxDHAceyZ",
    "outputId": "0c91d612-3b8f-488a-c9aa-df4064fda350"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-b3f5c7e8-fa40-4f84-9b28-9d6938f492ff\">\n",
       "    <div class=\"colab-df-container\">\n",
       "      <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>support</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>dressup</th>\n",
       "      <td>0.166667</td>\n",
       "      <td>0.750000</td>\n",
       "      <td>0.272727</td>\n",
       "      <td>4.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>makeup</th>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.181818</td>\n",
       "      <td>0.235294</td>\n",
       "      <td>11.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>money</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>12.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>netflix</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>12.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>trending</th>\n",
       "      <td>0.375000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.545455</td>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>accuracy</th>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.428571</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>macro avg</th>\n",
       "      <td>0.575000</td>\n",
       "      <td>0.553030</td>\n",
       "      <td>0.444029</td>\n",
       "      <td>42.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted avg</th>\n",
       "      <td>0.701389</td>\n",
       "      <td>0.428571</td>\n",
       "      <td>0.459893</td>\n",
       "      <td>42.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-b3f5c7e8-fa40-4f84-9b28-9d6938f492ff')\"\n",
       "              title=\"Convert this dataframe to an interactive table.\"\n",
       "              style=\"display:none;\">\n",
       "        \n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
       "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
       "  </svg>\n",
       "      </button>\n",
       "      \n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      flex-wrap:wrap;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "      <script>\n",
       "        const buttonEl =\n",
       "          document.querySelector('#df-b3f5c7e8-fa40-4f84-9b28-9d6938f492ff button.colab-df-convert');\n",
       "        buttonEl.style.display =\n",
       "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "        async function convertToInteractive(key) {\n",
       "          const element = document.querySelector('#df-b3f5c7e8-fa40-4f84-9b28-9d6938f492ff');\n",
       "          const dataTable =\n",
       "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                     [key], {});\n",
       "          if (!dataTable) return;\n",
       "\n",
       "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "            + ' to learn more about interactive tables.';\n",
       "          element.innerHTML = '';\n",
       "          dataTable['output_type'] = 'display_data';\n",
       "          await google.colab.output.renderOutput(dataTable, element);\n",
       "          const docLink = document.createElement('div');\n",
       "          docLink.innerHTML = docLinkHtml;\n",
       "          element.appendChild(docLink);\n",
       "        }\n",
       "      </script>\n",
       "    </div>\n",
       "  </div>\n",
       "  "
      ],
      "text/plain": [
       "              precision    recall  f1-score    support\n",
       "dressup        0.166667  0.750000  0.272727   4.000000\n",
       "makeup         0.333333  0.181818  0.235294  11.000000\n",
       "money          1.000000  0.500000  0.666667  12.000000\n",
       "netflix        1.000000  0.333333  0.500000  12.000000\n",
       "trending       0.375000  1.000000  0.545455   3.000000\n",
       "accuracy       0.428571  0.428571  0.428571   0.428571\n",
       "macro avg      0.575000  0.553030  0.444029  42.000000\n",
       "weighted avg   0.701389  0.428571  0.459893  42.000000"
      ]
     },
     "execution_count": 168,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class_report = metrics.classification_report(test_y, predictions, output_dict=True)\n",
    "report_df = pd.DataFrame(class_report).transpose()\n",
    "report_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "7hCtHjAUJEFM"
   },
   "source": [
    "- Common metrics for the performance of classification models\n",
    "\n",
    "![](https://qph.fs.quoracdn.net/main-qimg-00e243e8fb31fbd3b49034c48ec0a915)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3V9F5-Zr2AG1",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### Transformer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Lf4BpyZZl87h",
    "outputId": "764aac13-4bf9-49aa-a38a-366bedfdc6af"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "nvcc: NVIDIA (R) Cuda compiler driver\n",
      "Copyright (c) 2005-2020 NVIDIA Corporation\n",
      "Built on Mon_Oct_12_20:09:46_PDT_2020\n",
      "Cuda compilation tools, release 11.1, V11.1.105\n",
      "Build cuda_11.1.TC455_06.29190527_0\n"
     ]
    }
   ],
   "source": [
    "!nvcc --version"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "btwGwJYAUCBU"
   },
   "source": [
    "- Change `cuda111` to the correponding cuda version of your machine."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "nk2zMpIGlr2u"
   },
   "outputs": [],
   "source": [
    "!pip install -U spacy[cuda111,transformers]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "OZnUHGTkj_NG"
   },
   "outputs": [],
   "source": [
    "!python -m spacy download zh_core_web_trf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "id": "96Igjj5-gTqV"
   },
   "outputs": [],
   "source": [
    "import spacy_transformers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "id": "sRSlHHRiedSS"
   },
   "outputs": [],
   "source": [
    "spacy.prefer_gpu()\n",
    "nlp = spacy.load(\"zh_core_web_trf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "4QJIEEYerDAS",
    "outputId": "c0bf0478-d7cc-48ed-e0d5-427201c95b8f"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['transformer', 'tagger', 'parser', 'attribute_ruler', 'ner']"
      ]
     },
     "execution_count": 99,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nlp.pipe_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ibj9mtCDn6nS",
    "outputId": "59606a9b-e857-40cc-984d-6f04b70abe92"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "WordpieceBatch(strings=[['[CLS]', '宜', '家', '家', '居', '新', '店', '店', '店', '長', '的', '名', '字', '好', '長', '喔', '！', '[SEP]']], input_ids=array([[ 101, 2139, 2157, 2157, 2233, 3173, 2421, 2421, 2421, 7269, 4638,\n",
       "        1399, 2099, 1962, 7269, 1595, 8013,  102]], dtype=int32), attention_mask=array([[1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,\n",
       "        1., 1.]], dtype=float32), lengths=[18], token_type_ids=array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],\n",
       "      dtype=int32))"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text = \"宜家家居新店店店長的名字好長喔！\"\n",
    "doc = nlp(text)\n",
    "doc._.trf_data.wordpieces"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0bYM2Zf-UQyk"
   },
   "source": [
    "- The sample text includes 18 tokens, each in 768 dimensions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zMu2zjbGoucy",
    "outputId": "c9085fa1-3e2c-4d94-cafc-778b78d5f3bc"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1, 18, 768)"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "doc._.trf_data.tensors[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "vTeI9wK_sCZc",
    "outputId": "d6e0ed19-9c42-45a6-ec93-1248e91f65f2"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['trending', 'dressup', 'netflix', 'money', 'makeup']"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique_labels = train_df['forum_en'].unique().tolist()\n",
    "unique_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "id": "80ITsjJJo_AE"
   },
   "outputs": [],
   "source": [
    "train_df['tuples'] = train_df.apply(lambda row: (row['excerpt'], row['forum_en']), axis=1)\n",
    "train_data = train_df['tuples'].tolist()\n",
    "test_df['tuples'] = test_df.apply(lambda row: (row['excerpt'], row['forum_en']), axis=1)\n",
    "test_data = test_df['tuples'].tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "H4-aj3R-9CSB",
    "outputId": "ca97f330-b28b-4d8a-8d17-19b4fb6946e4"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('位於烏克蘭、歐洲最大的札波羅熱（Zaporizhzhia）核電廠在俄軍攻勢下起火，美國總統拜登稍早立即致電烏克蘭總統澤倫斯基（Volodymyr Zelensky），白宮將會持續密切關注核電廠火災狀況',\n",
       "  'trending'),\n",
       " ('最近一直看有沒有不錯的賣場，有人知道這家是不是正的嗎，謝謝🥺🥺', 'dressup'),\n",
       " ('2/13-2/15我踏上了一場說走就走的旅程，一個人三天兩夜的墾丁之旅，穿搭偏簡單 走黑白色系\\U0001f90d，（畢竟自己出門 不要帶太多東西比較好，雖然偏簡單 還是想記錄一下，如果有想自己出遊的人也可以做個參考～',\n",
       "  'dressup')]"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data[-3:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {
    "id": "iM7fhjwDu6bM"
   },
   "outputs": [],
   "source": [
    "from tqdm.auto import tqdm\n",
    "from spacy.tokens import DocBin\n",
    "\n",
    "def make_docs(data_list, unique_labels, dest_path):\n",
    "    nlp = spacy.blank('zh')\n",
    "    docs = []\n",
    "    for doc, label in tqdm(nlp.pipe(data_list, as_tuples=True), total=len(data_list)):\n",
    "        label_dict = {label: False for label in unique_labels}\n",
    "        doc.cats = label_dict\n",
    "        doc.cats[label] = True\n",
    "        docs.append(doc)\n",
    "    doc_bin = DocBin(docs=docs)\n",
    "    doc_bin.to_disk(dest_path)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 81,
     "referenced_widgets": [
      "e60a9cfc4b7748b5b53df301c4ed2c96",
      "a022c2c3fdb040e9bd2b5f86499ef12d",
      "88cf09ad12a5405d9649916cdedbbc73",
      "4c11909683a74344af573335c194d679",
      "8b0c9fd10f754442ba355eeea81b6386",
      "bff7110543ba40ab9ee4614c23fa73f9",
      "e3ee0ffd00544ba0af21e8a175b173aa",
      "68d1a81d31a44f16ae7f9bd17a0c2f51",
      "7d5ec8d7f8004caeba8d15c91be9dcab",
      "35042fa88e854be3b4d08f002e148159",
      "a971f7f377b24a649fa49f4646faa2d4",
      "fd1eb0585e944d5599594694bd055c97",
      "b2a8a0c2a98e47d7bca9720510fde9d4",
      "428506ae72254f8188986621634b0d2d",
      "8bc1782c6ca24192a715fd5c1e372e98",
      "1ba020a319cf4773bc951dc7e88f03e4",
      "dd70832bca6f4e33b436bb1ad431a65b",
      "4f331eb06d5244f7b206393892302b7c",
      "8f319eb7851647ad9c93e415ebee2e7e",
      "90204d21e2d04524a6f2570d5c29381f",
      "92e1966f42154300b5f80b121bba22a7",
      "d50ce50f118342f2b86c2e0786b4207e"
     ]
    },
    "id": "kVDF0GYjwTNE",
    "outputId": "a0a59f48-998a-4386-fb25-a863b8bf381f"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e60a9cfc4b7748b5b53df301c4ed2c96",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/98 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fd1eb0585e944d5599594694bd055c97",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/42 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "make_docs(train_data, unique_labels, \"train.spacy\")\n",
    "make_docs(test_data, unique_labels, \"test.spacy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "SYR6ijdD3TRX",
    "outputId": "ca7e0eff-0934-4600-fab6-274c1010370e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[38;5;4mℹ Generated config template specific for your use case\u001b[0m\n",
      "- Language: zh\n",
      "- Pipeline: textcat\n",
      "- Optimize for: accuracy\n",
      "- Hardware: GPU\n",
      "- Transformer: bert-base-chinese\n",
      "\u001b[38;5;2m✔ Auto-filled config with all values\u001b[0m\n",
      "\u001b[38;5;2m✔ Saved config\u001b[0m\n",
      "config.cfg\n",
      "You can now add your data and train your pipeline:\n",
      "python -m spacy train config.cfg --paths.train ./train.spacy --paths.dev ./dev.spacy\n"
     ]
    }
   ],
   "source": [
    "!python -m spacy init config ./config.cfg --lang zh --pipeline transformer,textcat --optimize accuracy --gpu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "UwCeyEbDMFeW",
    "outputId": "8282ddac-93ba-4558-c95c-8eadd5436263"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[paths]\n",
      "train = null\n",
      "dev = null\n",
      "vectors = null\n",
      "init_tok2vec = null\n",
      "\n",
      "[system]\n",
      "gpu_allocator = \"pytorch\"\n",
      "seed = 0\n",
      "\n",
      "[nlp]\n",
      "lang = \"zh\"\n",
      "pipeline = [\"transformer\",\"textcat\"]\n",
      "batch_size = 128\n",
      "disabled = []\n",
      "before_creation = null\n",
      "after_creation = null\n",
      "after_pipeline_creation = null\n",
      "\n",
      "[nlp.tokenizer]\n",
      "@tokenizers = \"spacy.zh.ChineseTokenizer\"\n",
      "segmenter = \"char\"\n",
      "\n",
      "[components]\n",
      "\n",
      "[components.textcat]\n",
      "factory = \"textcat\"\n",
      "scorer = {\"@scorers\":\"spacy.textcat_scorer.v1\"}\n",
      "threshold = 0.5\n",
      "\n",
      "[components.textcat.model]\n",
      "@architectures = \"spacy.TextCatEnsemble.v2\"\n",
      "nO = null\n",
      "\n",
      "[components.textcat.model.linear_model]\n",
      "@architectures = \"spacy.TextCatBOW.v2\"\n",
      "exclusive_classes = true\n",
      "ngram_size = 1\n",
      "no_output_layer = false\n",
      "nO = null\n",
      "\n",
      "[components.textcat.model.tok2vec]\n",
      "@architectures = \"spacy-transformers.TransformerListener.v1\"\n",
      "grad_factor = 1.0\n",
      "pooling = {\"@layers\":\"reduce_mean.v1\"}\n",
      "upstream = \"*\"\n",
      "\n",
      "[components.transformer]\n",
      "factory = \"transformer\"\n",
      "max_batch_items = 4096\n",
      "set_extra_annotations = {\"@annotation_setters\":\"spacy-transformers.null_annotation_setter.v1\"}\n",
      "\n",
      "[components.transformer.model]\n",
      "@architectures = \"spacy-transformers.TransformerModel.v3\"\n",
      "name = \"bert-base-chinese\"\n",
      "mixed_precision = false\n",
      "\n",
      "[components.transformer.model.get_spans]\n",
      "@span_getters = \"spacy-transformers.strided_spans.v1\"\n",
      "window = 128\n",
      "stride = 96\n",
      "\n",
      "[components.transformer.model.grad_scaler_config]\n",
      "\n",
      "[components.transformer.model.tokenizer_config]\n",
      "use_fast = true\n",
      "\n",
      "[components.transformer.model.transformer_config]\n",
      "\n",
      "[corpora]\n",
      "\n",
      "[corpora.dev]\n",
      "@readers = \"spacy.Corpus.v1\"\n",
      "path = ${paths.dev}\n",
      "max_length = 0\n",
      "gold_preproc = false\n",
      "limit = 0\n",
      "augmenter = null\n",
      "\n",
      "[corpora.train]\n",
      "@readers = \"spacy.Corpus.v1\"\n",
      "path = ${paths.train}\n",
      "max_length = 0\n",
      "gold_preproc = false\n",
      "limit = 0\n",
      "augmenter = null\n",
      "\n",
      "[training]\n",
      "accumulate_gradient = 3\n",
      "dev_corpus = \"corpora.dev\"\n",
      "train_corpus = \"corpora.train\"\n",
      "seed = ${system.seed}\n",
      "gpu_allocator = ${system.gpu_allocator}\n",
      "dropout = 0.1\n",
      "patience = 1600\n",
      "max_epochs = 0\n",
      "max_steps = 20000\n",
      "eval_frequency = 200\n",
      "frozen_components = []\n",
      "annotating_components = []\n",
      "before_to_disk = null\n",
      "\n",
      "[training.batcher]\n",
      "@batchers = \"spacy.batch_by_padded.v1\"\n",
      "discard_oversize = true\n",
      "size = 2000\n",
      "buffer = 256\n",
      "get_length = null\n",
      "\n",
      "[training.logger]\n",
      "@loggers = \"spacy.ConsoleLogger.v1\"\n",
      "progress_bar = false\n",
      "\n",
      "[training.optimizer]\n",
      "@optimizers = \"Adam.v1\"\n",
      "beta1 = 0.9\n",
      "beta2 = 0.999\n",
      "L2_is_weight_decay = true\n",
      "L2 = 0.01\n",
      "grad_clip = 1.0\n",
      "use_averages = false\n",
      "eps = 0.00000001\n",
      "\n",
      "[training.optimizer.learn_rate]\n",
      "@schedules = \"warmup_linear.v1\"\n",
      "warmup_steps = 250\n",
      "total_steps = 20000\n",
      "initial_rate = 0.00005\n",
      "\n",
      "[training.score_weights]\n",
      "cats_score = 1.0\n",
      "cats_score_desc = null\n",
      "cats_micro_p = null\n",
      "cats_micro_r = null\n",
      "cats_micro_f = null\n",
      "cats_macro_p = null\n",
      "cats_macro_r = null\n",
      "cats_macro_f = null\n",
      "cats_macro_auc = null\n",
      "cats_f_per_type = null\n",
      "cats_macro_auc_per_type = null\n",
      "\n",
      "[pretraining]\n",
      "\n",
      "[initialize]\n",
      "vectors = ${paths.vectors}\n",
      "init_tok2vec = ${paths.init_tok2vec}\n",
      "vocab_data = null\n",
      "lookups = null\n",
      "before_init = null\n",
      "after_init = null\n",
      "\n",
      "[initialize.components]\n",
      "\n",
      "[initialize.tokenizer]\n",
      "pkuseg_model = null\n",
      "pkuseg_user_dict = \"default\""
     ]
    }
   ],
   "source": [
    "!cat config.cfg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {
    "id": "bb0OjTdSyODb"
   },
   "outputs": [],
   "source": [
    "!mkdir output"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YGKC-FfJV1Z_"
   },
   "source": [
    "- Abort the execution of the following cell if it runs for too long."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "q7xchz0yxj2u",
    "outputId": "dafe13d5-37d4-45dc-cd2b-8f3c23259b7c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2022-03-18 13:32:23,319] [DEBUG] Config overrides from CLI: ['paths.train', 'paths.dev']\n",
      "\u001b[38;5;4mℹ Saving to output directory: output\u001b[0m\n",
      "\u001b[38;5;4mℹ Using GPU: 0\u001b[0m\n",
      "\u001b[1m\n",
      "=========================== Initializing pipeline ===========================\u001b[0m\n",
      "[2022-03-18 13:32:26,310] [INFO] Set up nlp object from config\n",
      "[2022-03-18 13:32:26,323] [DEBUG] Loading corpus from path: test.spacy\n",
      "[2022-03-18 13:32:26,324] [DEBUG] Loading corpus from path: train.spacy\n",
      "[2022-03-18 13:32:26,324] [INFO] Pipeline: ['transformer', 'textcat']\n",
      "[2022-03-18 13:32:26,330] [INFO] Created vocabulary\n",
      "[2022-03-18 13:32:26,332] [INFO] Finished initializing nlp object\n",
      "Some weights of the model checkpoint at bert-base-chinese were not used when initializing BertModel: ['cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.dense.weight', 'cls.predictions.decoder.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.seq_relationship.bias', 'cls.seq_relationship.weight', 'cls.predictions.bias']\n",
      "- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "[2022-03-18 13:32:36,584] [INFO] Initialized pipeline components: ['transformer', 'textcat']\n",
      "\u001b[38;5;2m✔ Initialized pipeline\u001b[0m\n",
      "\u001b[1m\n",
      "============================= Training pipeline =============================\u001b[0m\n",
      "[2022-03-18 13:32:36,598] [DEBUG] Loading corpus from path: test.spacy\n",
      "[2022-03-18 13:32:36,599] [DEBUG] Loading corpus from path: train.spacy\n",
      "[2022-03-18 13:32:36,650] [DEBUG] Removed existing output directory: output/model-best\n",
      "[2022-03-18 13:32:36,704] [DEBUG] Removed existing output directory: output/model-last\n",
      "\u001b[38;5;4mℹ Pipeline: ['transformer', 'textcat']\u001b[0m\n",
      "\u001b[38;5;4mℹ Initial learn rate: 0.0\u001b[0m\n",
      "E    #       LOSS TRANS...  LOSS TEXTCAT  CATS_SCORE  SCORE \n",
      "---  ------  -------------  ------------  ----------  ------\n",
      "  0       0           0.00          0.80        0.00    0.00\n",
      " 66     200           0.00         36.54       94.98    0.95\n",
      "\n",
      "Aborted!\n",
      "It took 471.73191380500793 secs to train the model.\n"
     ]
    }
   ],
   "source": [
    "start = time.time()\n",
    "!python -m spacy train ./config.cfg --output ./output --paths.train ./train.spacy --paths.dev ./test.spacy --gpu-id 0 --verbose\n",
    "end = time.time()\n",
    "print(f\"It took {end-start} secs to train the model.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sFAIF8pzWDKx"
   },
   "source": [
    "- Load the model that we just trained."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "0opmsOeZRLCk"
   },
   "outputs": [],
   "source": [
    "mynlp = spacy.load(\"./output/model-last\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "qZgSQ38zNYyG",
    "outputId": "db476e07-6c8b-4576-f16b-7fe7e61d1a5c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[139, 97, 134, 110, 81, 26, 35, 32, 84, 95, 25, 69, 40, 28, 125, 133, 99, 117, 11, 46, 51, 101, 136, 121, 37, 21, 89, 119, 118, 29, 96, 52, 45, 106, 92, 62, 130, 82, 33, 105, 111, 120]\n"
     ]
    }
   ],
   "source": [
    "test_indexes = test_df.index.tolist()\n",
    "print(test_indexes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_3AQPuquWIta"
   },
   "source": [
    "- Feel free to change the value of `docid` to any number in `test_indexes`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "GHUvvCH4NxRH",
    "outputId": "3c3c2d47-dc08-458c-b57b-11e2a5bb6b35"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sample text: \n",
      "#重雷，因為版上都是讚美的聲音…，其實幾位主角性格都很討喜，只是太多人大大大推了 個人看了反而有點小失望，第一集開頭用2020疫情和1997金融海嘯來提到，「時代影響了很多人」也覺得很有共鳴，因為我們\n",
      "True category: \n",
      "netflix\n"
     ]
    }
   ],
   "source": [
    "docid = 139\n",
    "sample_text = test_df.loc[docid, 'excerpt']\n",
    "sample_cat = test_df.loc[docid, 'forum_en']\n",
    "print(f\"Sample text: \\n{sample_text}\")\n",
    "print(f\"True category: \\n{sample_cat}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "o90PCYO60bKb",
    "outputId": "dd89fe34-f894-41dd-b371-a768d847c198"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Category probabilities:\n",
      "{'trending': 0.002882840344682336, 'dressup': 0.0029310560785233974, 'netflix': 0.9875414967536926, 'money': 0.0030496492981910706, 'makeup': 0.003594897221773863}\n",
      "Predicted category: netflix\n"
     ]
    }
   ],
   "source": [
    "doc = mynlp(sample_text)\n",
    "max_cat = max(doc.cats, key=doc.cats.get)\n",
    "print(f\"Category probabilities:\\n{doc.cats}\")\n",
    "print(f\"Predicted category: {max_cat}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "GlDxtXEgWTkv"
   },
   "source": [
    "- Feel free to revise the sentence stored in `text`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "OPcTZUG4RveK",
    "outputId": "412a4fb4-f95b-4929-9bc8-3edf4768c670"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Category probabilities:\n",
      "{'trending': 0.0022079325281083584, 'dressup': 0.9901031255722046, 'netflix': 0.002575943013653159, 'money': 0.002191048115491867, 'makeup': 0.0029218546114861965}\n",
      "Predicted category: dressup\n"
     ]
    }
   ],
   "source": [
    "text = \"今年似乎又開始流行連身裙了呢\"\n",
    "doc = mynlp(text)\n",
    "max_cat = max(doc.cats, key=doc.cats.get)\n",
    "print(f\"Category probabilities:\\n{doc.cats}\")\n",
    "print(f\"Predicted category: {max_cat}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "24YR5mu4Riba"
   },
   "outputs": [],
   "source": [
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "S8Sj3gf8PPQy",
    "outputId": "29d07918-a9c8-45bb-914e-148079c0f012"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'cats_f_per_type': {'dressup': {'f': 0.8888888889, 'p': 0.8, 'r': 1.0},\n",
       "  'makeup': {'f': 0.9, 'p': 1.0, 'r': 0.8181818182},\n",
       "  'money': {'f': 0.96, 'p': 0.9230769231, 'r': 1.0},\n",
       "  'netflix': {'f': 1.0, 'p': 1.0, 'r': 1.0},\n",
       "  'trending': {'f': 1.0, 'p': 1.0, 'r': 1.0}},\n",
       " 'cats_macro_auc': 0.9899175113,\n",
       " 'cats_macro_auc_per_type': 0.0,\n",
       " 'cats_macro_f': 0.9497777778,\n",
       " 'cats_macro_p': 0.9446153846,\n",
       " 'cats_macro_r': 0.9636363636,\n",
       " 'cats_micro_f': 0.9523809524,\n",
       " 'cats_micro_p': 0.9523809524,\n",
       " 'cats_micro_r': 0.9523809524,\n",
       " 'cats_score': 0.9497777778,\n",
       " 'cats_score_desc': 'macro F',\n",
       " 'textcat_loss': 36.5435999353,\n",
       " 'transformer_loss': 0.0006216127}"
      ]
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_meta_path = \"./output/model-best/meta.json\"\n",
    "with open(model_meta_path) as json_file:\n",
    "    trained_metrics = json.load(json_file)\n",
    "trained_metrics['performance'] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "id": "7NMeFWm5IU34"
   },
   "outputs": [],
   "source": [
    "import random\n",
    "\n",
    "def show_test():\n",
    "    idx = random.choice(test_indexes)\n",
    "    text = test_df.at[idx, 'excerpt']\n",
    "    label = test_df.at[idx, 'forum_en']\n",
    "    predicted_proba = mynlp(text).cats\n",
    "    predicted_cat = max(predicted_proba, key=predicted_proba.get)\n",
    "    print(f\"Index in test data: {idx}\")\n",
    "    print(f\"Text: {text}\")\n",
    "    print(f\"True category: {label}\")\n",
    "    print(f\"Category probabilities:\\n{predicted_proba}\")\n",
    "    print(f\"Predicted category: {predicted_cat}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "UhMaxlewJ41G",
    "outputId": "f8693537-b085-4b55-9898-bc9cb55cc36b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index in test data: 25\n",
      "Text: 有人知道這是什麼牌子嗎\n",
      "True category: dressup\n",
      "Category probabilities:\n",
      "{'trending': 0.0024371894542127848, 'dressup': 0.9893414974212646, 'netflix': 0.0024927842896431684, 'money': 0.0022952246945351362, 'makeup': 0.003433354664593935}\n",
      "Predicted category: dressup\n"
     ]
    }
   ],
   "source": [
    "show_test()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "RCplR7z9Sn26",
    "outputId": "581cc676-327b-4446-daa7-403e37172669"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index in test data: 32\n",
      "Text: 因為自己本身是敏感肌，所以用保養品都會很小心翼翼看成分，目前覺得醫美牌大致上幾乎都蠻安全的，算是可放心服用～另外味道也是一門學問，基本上香味太濃郁妖豔的，擦上去就是立馬泛紅或刺刺的，百試不厭耶！，而且\n",
      "True category: makeup\n",
      "Category probabilities:\n",
      "{'trending': 0.0046507827937603, 'dressup': 0.005224997643381357, 'netflix': 0.005102104041725397, 'money': 0.005229421425610781, 'makeup': 0.9797927141189575}\n",
      "Predicted category: makeup\n"
     ]
    }
   ],
   "source": [
    "show_test()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "if3IUBf-SqmY",
    "outputId": "3fef5d42-e0bc-4555-9460-c0c877c3aade"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index in test data: 92\n",
      "Text: 2022.03.03，當日收益：151.1178，本金：50w，註記：無，程式使用API串聯，幣安，火幣，kraken，ok等一線交易所進行交易，資金皆在自己的交易所帳戶內️️，Instagram：加密\n",
      "True category: money\n",
      "Category probabilities:\n",
      "{'trending': 0.003693468403071165, 'dressup': 0.003174132201820612, 'netflix': 0.0034728439059108496, 'money': 0.9856423735618591, 'makeup': 0.004017228726297617}\n",
      "Predicted category: money\n"
     ]
    }
   ],
   "source": [
    "show_test()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6tCEKKFNTKRT"
   },
   "source": [
    "- With powerful techniques like deep learning and transfer learning, it's sometimes not easy to explain to customers/clients/bosses why your models make good predictions.\n",
    "\n",
    "![](https://github.com/howard-haowen/NLP-demos/raw/main/img/explain-models.JPG)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1xzDmyLgBHmX",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# References"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "a11-DzEqjaJf",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Articles\n",
    "\n",
    "- [An Introduction to Machine Learning](https://monkeylearn.com/machine-learning/)\n",
    "- [Top Natural Language\n",
    "Processing Applications\n",
    "in Business](https://www.accenture.com/_acnmedia/pdf-106/accenture-unlocking-value-unstructured-data.pdf)\n",
    "- [Symbolic, Distributed, and Distributional Representations for Natural Language Processing in the Era of Deep Learning: A Survey](https://www.frontiersin.org/articles/10.3389/frobt.2019.00153/full)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zc49uC4hBB3_",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Books\n",
    "\n",
    "<div>\n",
    "<img src=\"https://i.gr-assets.com/images/S/compressed.photo.goodreads.com/books/1630086235l/58870327._SX318_.jpg\" width:50%>\n",
    "<img src=\"https://i.gr-assets.com/images/S/compressed.photo.goodreads.com/books/1628096299l/55711023._SX318_.jpg\" width:50%>\n",
    "</div>"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "colab": {
   "collapsed_sections": [],
   "include_colab_link": true,
   "name": "NTNU_talk.ipynb",
   "provenance": []
  },
  "jupytext": {
   "formats": "ipynb,auto:percent"
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.2"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "1ba020a319cf4773bc951dc7e88f03e4": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "35042fa88e854be3b4d08f002e148159": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "428506ae72254f8188986621634b0d2d": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_8f319eb7851647ad9c93e415ebee2e7e",
      "max": 42,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_90204d21e2d04524a6f2570d5c29381f",
      "value": 42
     }
    },
    "4c11909683a74344af573335c194d679": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_35042fa88e854be3b4d08f002e148159",
      "placeholder": "​",
      "style": "IPY_MODEL_a971f7f377b24a649fa49f4646faa2d4",
      "value": " 98/98 [00:00&lt;00:00, 666.47it/s]"
     }
    },
    "4f331eb06d5244f7b206393892302b7c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "68d1a81d31a44f16ae7f9bd17a0c2f51": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "7d5ec8d7f8004caeba8d15c91be9dcab": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "88cf09ad12a5405d9649916cdedbbc73": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_68d1a81d31a44f16ae7f9bd17a0c2f51",
      "max": 98,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_7d5ec8d7f8004caeba8d15c91be9dcab",
      "value": 98
     }
    },
    "8b0c9fd10f754442ba355eeea81b6386": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "8bc1782c6ca24192a715fd5c1e372e98": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_92e1966f42154300b5f80b121bba22a7",
      "placeholder": "​",
      "style": "IPY_MODEL_d50ce50f118342f2b86c2e0786b4207e",
      "value": " 42/42 [00:00&lt;00:00, 494.93it/s]"
     }
    },
    "8f319eb7851647ad9c93e415ebee2e7e": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "90204d21e2d04524a6f2570d5c29381f": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "92e1966f42154300b5f80b121bba22a7": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "a022c2c3fdb040e9bd2b5f86499ef12d": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_bff7110543ba40ab9ee4614c23fa73f9",
      "placeholder": "​",
      "style": "IPY_MODEL_e3ee0ffd00544ba0af21e8a175b173aa",
      "value": "100%"
     }
    },
    "a971f7f377b24a649fa49f4646faa2d4": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "b2a8a0c2a98e47d7bca9720510fde9d4": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_dd70832bca6f4e33b436bb1ad431a65b",
      "placeholder": "​",
      "style": "IPY_MODEL_4f331eb06d5244f7b206393892302b7c",
      "value": "100%"
     }
    },
    "bff7110543ba40ab9ee4614c23fa73f9": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "d50ce50f118342f2b86c2e0786b4207e": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "dd70832bca6f4e33b436bb1ad431a65b": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "e3ee0ffd00544ba0af21e8a175b173aa": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "e60a9cfc4b7748b5b53df301c4ed2c96": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_a022c2c3fdb040e9bd2b5f86499ef12d",
       "IPY_MODEL_88cf09ad12a5405d9649916cdedbbc73",
       "IPY_MODEL_4c11909683a74344af573335c194d679"
      ],
      "layout": "IPY_MODEL_8b0c9fd10f754442ba355eeea81b6386"
     }
    },
    "fd1eb0585e944d5599594694bd055c97": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_b2a8a0c2a98e47d7bca9720510fde9d4",
       "IPY_MODEL_428506ae72254f8188986621634b0d2d",
       "IPY_MODEL_8bc1782c6ca24192a715fd5c1e372e98"
      ],
      "layout": "IPY_MODEL_1ba020a319cf4773bc951dc7e88f03e4"
     }
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}