workshop_7_aspect_based_sentiment_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Unsupervised approach for Aspect Based Sentiment Analysis on Restaurant Reviews"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We introduce simplest possible approach for unsupervised aspect based sentiment analysis. Here we use word2vec embeddings from spaCy for matching candidate terms from sentences to aspects (classification of aspect terms) and dependency tree parsing using spaCy to detect references of opinion words to aspect terms. Additional list of positive and negative opinion words list is used for classifying sentiment and further analysis of dependencies is used for polarity detection (negations, verbs, adjectives of opinion words, etc).  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Install Requirements"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !pip install pandas==1.0.3 \\\n",
    "# scikit-learn==0.21.3 \\\n",
    "# scikit-multilearn==0.2.0 \\\n",
    "# neuralcoref==4.0.0 \\\n",
    "# spacy==2.1.0 \\\n",
    "# plotly==4.6.0 \\\n",
    "# gensim==3.8.0 \\\n",
    "# matplotlib==3.1.3 \\\n",
    "# six==1.12.0 \\\n",
    "# numpy==1.16.5 \\\n",
    "# nltk==3.4.5 \\\n",
    "# cython"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Download large english model which has word embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !python -m spacy download en_core_web_lg"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "import json\n",
    "import re\n",
    "import spacy\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Retrieve dataset and parse"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dataset file is xml format. We will parse them using xml library"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import xml.etree.ElementTree as ET"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Download dataset file from github source link"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists('absa_data'):\n",
    "    !mkdir absa_data\n",
    "\n",
    "if not os.path.exists('absa_data/Restaurants_Train.xml'):\n",
    "    !cd absa_data && wget https://raw.githubusercontent.com/LFCapollo/Aspect-Based-Sentiment-Analysis/master/data/Restaurants_Train.xml"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "parsing of xml to extract sentences, aspect terms, aspect categories and sentiment polarities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parse_xml_data(root):\n",
    "    \"\"\"Parses xml data to extract aspect terms and aspect categories from each sentence\"\"\"\n",
    "    \n",
    "    reviews = []\n",
    "\n",
    "    for sentence in root.findall(\"sentence\"):\n",
    "        entry = {}\n",
    "        aspect_terms = []\n",
    "        aspect_categories = []\n",
    "        \n",
    "        if sentence.find(\"aspectTerms\"):\n",
    "            for aspect_term in sentence.find(\"aspectTerms\").findall(\"aspectTerm\"):\n",
    "                aspect_terms.append((aspect_term.get(\"term\"), aspect_term.get(\"polarity\")))\n",
    "                \n",
    "        if sentence.find(\"aspectCategories\"):\n",
    "            for aspect_category in sentence.find(\"aspectCategories\").findall(\"aspectCategory\"):\n",
    "                aspect_categories.append((aspect_category.get(\"category\"), aspect_category.get(\"polarity\")))\n",
    "                \n",
    "        entry[\"text\"] = sentence[0].text\n",
    "        entry[\"terms\"] = aspect_terms\n",
    "        entry[\"aspects\"] = aspect_categories\n",
    "        \n",
    "        reviews.append(entry)\n",
    "\n",
    "    reviews_df = pd.DataFrame(reviews)\n",
    "    \n",
    "    return reviews_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "tree = ET.parse('absa_data/Restaurants_Train.xml')\n",
    "root = tree.getroot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = parse_xml_data(root)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>text</th>\n",
       "      <th>terms</th>\n",
       "      <th>aspects</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>But the staff was so horrible to us.</td>\n",
       "      <td>[(staff, negative)]</td>\n",
       "      <td>[(service, negative)]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>To be completely fair, the only redeeming fact...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[(food, positive), (anecdotes/miscellaneous, n...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>The food is uniformly exceptional, with a very...</td>\n",
       "      <td>[(food, positive), (kitchen, positive), (menu,...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Where Gabriela personaly greets you and recomm...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(service, positive)]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>For those that go once and don't enjoy it, all...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(anecdotes/miscellaneous, positive)]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  \\\n",
       "0               But the staff was so horrible to us.   \n",
       "1  To be completely fair, the only redeeming fact...   \n",
       "2  The food is uniformly exceptional, with a very...   \n",
       "3  Where Gabriela personaly greets you and recomm...   \n",
       "4  For those that go once and don't enjoy it, all...   \n",
       "\n",
       "                                               terms  \\\n",
       "0                                [(staff, negative)]   \n",
       "1                                 [(food, positive)]   \n",
       "2  [(food, positive), (kitchen, positive), (menu,...   \n",
       "3                                                 []   \n",
       "4                                                 []   \n",
       "\n",
       "                                             aspects  \n",
       "0                              [(service, negative)]  \n",
       "1  [(food, positive), (anecdotes/miscellaneous, n...  \n",
       "2                                 [(food, positive)]  \n",
       "3                              [(service, positive)]  \n",
       "4              [(anecdotes/miscellaneous, positive)]  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# EDA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We do simple exploratory data analysis. Data is much cleaner than usual but some additional cleaning can be applied for improving overall results."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load spaCy english large model (en_core_web_lg - contains word embeddings also)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "load spaCy english model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "nlp = spacy.load('en_core_web_lg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Count Records"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We don't have too much data. Just several thousand records"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3044"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show sample values per column"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just to have a look at sample values from each column. As we see we have texts (one or more sentences, aspect terms and aspects with polarity values for both)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "==================================================\n",
      "Column: TEXT\n",
      "==================================================\n",
      "Values: ['Oh yes, and they lie on the phone, claiming they have seating in the garden, then of course the seats are not available.', 'So my husband made up an excuse so we can leave.', 'My friends and I experienced amazing cheese and a delicious, new summer menu at Artisanal last night.', 'For those prices and the dressed up atmosphere you expect more and should get more.', \"I'm no food critic, but I'd like to think I have a tiny bit of experience under my belt having lived in NY for the last 11 years.\", 'If you go here - your focus should be the Tamarind Margaritas.', 'I cannot imagine better Indian food in all of the city.', 'A glass of Leaping Lizard, a glass of prosecco, and the mussels had everything happy.', 'The staff makes you feel at home, the food is great and the atmosphere is WONDERFUL!', 'Appetizers are somewhere around $7 each and the main dishes are between $11 and $16.']\n",
      "--------------------------------------------------\n",
      "==================================================\n",
      "Column: TERMS\n",
      "==================================================\n",
      "Values: [[], [('Decor', 'positive'), ('food', 'positive'), ('wine list', 'positive'), ('food', 'neutral')], [], [('service', 'positive'), ('atmosphere', 'positive')], [], [('chef', 'negative'), ('staff', 'negative')], [('fish', 'negative'), ('rice', 'negative')], [('Tuscan cuisine', 'conflict'), ('Neapolitan fare', 'positive')], [('spot lights', 'negative')], [('price', 'negative')]]\n",
      "--------------------------------------------------\n",
      "==================================================\n",
      "Column: ASPECTS\n",
      "==================================================\n",
      "Values: [[('anecdotes/miscellaneous', 'neutral')], [('anecdotes/miscellaneous', 'positive')], [('food', 'neutral')], [('price', 'positive'), ('food', 'positive')], [('anecdotes/miscellaneous', 'negative')], [('ambience', 'negative'), ('price', 'positive')], [('food', 'positive'), ('ambience', 'positive')], [('food', 'positive')], [('service', 'positive')], [('anecdotes/miscellaneous', 'positive')]]\n",
      "--------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "for column in df.columns:\n",
    "    print('='*50)\n",
    "    print(f'Column: {column.upper()}')\n",
    "    print('='*50)\n",
    "    print(f'Values: {df[column].sample(10).values.tolist()}')\n",
    "    print('-'*50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Aspects"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have just 5 aspect that we need to analyse. (we see later that 'anecdotes/miscellaneous' is kind of hard to analyse with simplest approach and we skip that from evaluation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'ambience', 'anecdotes/miscellaneous', 'food', 'price', 'service'}"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "set(df.aspects.apply(lambda x: [item[0] for item in x]).sum())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Aspects distribution per text/sentence"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We have mostly one aspect (sometimes 2 and 3) per text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWsAAAEICAYAAACZJtWMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAWPklEQVR4nO3df7RdZX3n8feXBORHmAgkpRAwAVQEcUSbUh0pi4UomGhpV1dbnYqgdKitM2KLg3GBDu1gTW2l1VZrsdKAUpBVdapQSqmFUsaCXpwgYEQCRjEi4VciCLUC3/njea7ZOZxzf4RzcvPkvl9rnZVz9j77Od/n2Xt/zj577wuRmUiStm87zXQBkqTJGdaS1ADDWpIaYFhLUgMMa0lqgGEtSQ3Y7sM6Ij4WEe8ZUlvPiYhHI2JOfX1dRPzGMNqu7V0VEacMq71pfO55EfFARHx/W392PxGREfHcGfrsQyNidUQ8EhFvn4kapFGY0bCOiHUR8XjdsTZGxJci4q0R8ZO6MvOtmfm/p9jW8RO9JzO/k5nzMvPJIdR+bkR8qqf912TmRc+07WnW8RzgTODwzPzpPvOPreH50Z7pN0TEqduozG3pLODazNwzMz886E0RsSoinoiI/bZhbYPqOG8ma9iWIuLUiLhhpusAiIhd6kHOvGEeuPXLhmHYHo6sX5eZewKLgZXAu4BPDPtDImLusNvcTjwHeDAzN0zwnh8CJ0fEkm1S0ZBs5TpbDNw+Sbt7AL8MbALeuBWfscMZ/7U5yxwDrM7MR2e6kCnJzBl7AOuA43umHQU8BRxRX68CzqvPFwBXABuBh4B/pXzhfLIu8zjwKOXoagmQwGnAd4DrO9Pm1vauA94PfBn4AfB3wN513rHAd/vVC5wI/Afw4/p5t3Ta+436fCfgHODbwAbgYmB+nTdexym1tgeAsycYp/l1+ftre+fU9o+vfX6q1rGqz7LHAt8F/gz46870G4BT6/NzgU915vUbp/OAL9XP+QKwD3BJHbevAEs6yyfwduDu2rc/AnbqzH8LsAZ4GLgaWNyz7NuAO4FvDRiPX6AE8sZa22F1+j8DTwL/Xut8/oDl3wTcA5wB3NZn+xur/boPOL9nTE4HvgfcC7yzs9xOwArgLuBB4HLqtlTnH13Hb2P97FNrWz+mbEuPAl+o730XsB54BLgDeOWAfqwCPgZcU9/7Lz1j+YI676Hazq/2LPsXwN9TvsyP79P+qXUdPgJ8C/j1aazDt9Z1uBH4CBDAYXXdPFn7u7G+/1nAH1P2hftqn3br2X7PpOxH9wJv7nzWbsAHKfvFJsp2Pb7syzpjfgtwbE//zgd+F3hfz3bz5xONH7ALsBr4H/X1HOD/Au9lQDYMJS9HFcRT+vA+YV2nfwf4rc5GNR7W768rcuf6+Hkg+rXF5p3rYmCPulLHp3VDaD1wRH3PZ6ihxQRh3S/gOu2Nh/VbgLXAwcA84LPAJ3tq+3it68XAj6ih02c8LqZ8kexZl/0mcNqgOnuWPZaysf80JYAOrdOnG9ZrgUMoXxxfrzUcD8yt9f11z856LbA35cj/m51xOam2dVhd9hzgSz3LXlOX3a1Pf55PCZdX1W3grNreLr3rYIIx+SLwAWBf4AngZzrz/g04uT6fB7ysZ0wurdvKiyhfnuPbwxnAjcABlPD5S+DSOm8xJfDeUGveBziyd/uurw+lhPn+nc89ZEA/VtV2j6mf+SHghjpvj9rOm+s4v4TyxXl4Z9lNwCsoXzS79rS9B1tuL/sBL5zGOrwCeHZd//cDJ9Z5p47X2Hn/nwCfr+t8T8rBwPs72+8TwO/XsVsGPAbsVed/pK7zRZTQ/C91LBZRvjSX1f69qr5e2Pncb3T6dx2d7WYK43cE5YvqMODsuu7nDMqGoeTlsBuc1ocPDusbqUeabBnWv08JredO1habd66DJwmhlZ35h1O+FefwzMP6i8Bv9+yEP64rfryOAzrzvwy8vk+/5tSaDu9M+03gus7GPGlY1+cfAD5dn083rM/uzP8gcFXn9esoPye7O+uJnde/DXyxPr+K+kVTX+9E2fkWd5Y9boL+vAe4vGf59dSjJiYJa0p4PMXmsLwa+FBn/vXA7wELepYbH5MXdKZ9APhEfb6GzhEwJdzG1/e7gc8NqGcVW4b1cylHkMcDO0+y/6wCLuu8nkc5QjwQ+DXgX3ve/5fA/+ose/EEbe9BOSL9ZXq+NKe4Do/uzL8cWFGfn0onrClH3D+k84UEvJz6q4qy/T4+vi3WaRsoR8071Xkv7lP/u6gHR51pVwOn1OeHAGv77bv19YTjV1+fSTnifhh4Xmf6uYwgrLeHc9b9LKL89Oj1R5Rv9H+MiLsjYsUU2rpnGvO/Tfn2XjClKie2f22v2/ZcytHcuO7dG49RdrZeC2pNvW0t2oqa/hA4ISJevBXL3td5/nif1721947r/vX5YuBD9YLy+OmsYMv+TLTOthjXzHyqvn+q43EysCYzV9fXlwD/NSJ2rq9Poxy9fyMivhIRr51Gvz7X6dcaSnDuSwnPu6ZSXGauBd5B2eE3RMRlEbH/BIv8pJ4s514fqjUtBn5uvJ5a069TfmH160tvHT+kBNZbgXsj4sqIeEGnr5Otw6ls2wALgd2Bmzvt/UOdPu7BzHyiT3sLgF3pP7aLgV/p6f/RlC9RKEfcVw3qP1Mbv4vq+/4+M++coK2h2O7COiJ+lrLSn3bFODMfycwzM/NgynnL342IV47PHtDkoOnjDuw8fw7laOgByrf97p265rDlBjRZu9+jrMhu20+wZchNxQO1pt621k+zHTLzQeBPgd67a7boK1tukFurd1y/V5/fA/xmZj6789gtM7/ULXWCdrcY14iI+llTHY83AQdHxPfrrY7nU3b6ZQCZeWdmvgH4KcqX29/WC5JT6ddrevq1a2aur/MOGVDP0/qamX+TmUfXfmatY5Cf1BMR8yinEr5XP/NfeuqZl5m/NdFn99RxdWa+ihJw36Ccthvv62TrcGCzPa8foHzZv7DT1vzMHBTuvcv+O/3H9h7KkXW3xj0yc2Wdv4xyvn5QXVMZv49STvecEBFHT9DWUGw3YR0R/6kexVxG+Qlxa5/3vDYinlt30E2UI5en6uz7KOeHp+uNEXF4ROxOOc3yt1lu7fsmsGtELK9HXedQzoWNuw9Y0r3NsMelwO9ExEF1J/oDyimIJwa8v69ay+XA+yJiz4hYTLkosrW3Bp1POa93WGfaauCYeh/6fMrP9mfqf0bEXhFxIOV87qfr9I8B746IFwJExPyI+JVptHs5sDwiXlnXy5mU8/2TBkVEvJyyYx8FHFkfRwB/QwlxIuKNEbGwHrFvrIs+1WnmPRGxe63/zT39el9dP0TEwog4qc67BDg+In41IuZGxD4RcWSdt8V2W+8TPy4inkUJovELyIMsi4ijI2IXypfwjZl5DyVEnh8RJ0fEzvXxsxFx2ARtdcdq34g4qX5R/YhysWy8jmeyDu8DDqj1jv8y+jjwJxHxU7W9RRFxwmQN1WUvBM6PiP0jYk5EvLyO3aeA10XECXX6rvU21gPqvn4U5bpKt65ufkw4fhFxMvAzlNM6bwcuqvv5eFsTZcPWGfZ5lek8KOeAH6dcJNlEubjzNuqJ+uw5pwf8Tl3mh5SLZu/pvO8kyoXJjcA76Tnvmluedxx0N8gX6JyrrCviXso5sney5TnrfShH/w8DX+20170b5L2Ub+j7KRvPXv3q6F22zzjtVZe/v7b3XurdFUzjnHVn2ln180/tTPtIHbu1wH/rM07d83nn0bnzhHJ+tXv+L9l8N8iDlHPc3XV6MnBrHfN7gAt7ln3aNYme+n+JcpFzE+UOiBdOcRw/Bnymz/SjKIG0dx3nDZRwuh34xZ51Nn43yPeBszpt7ET5Er2Dsj3fBfxBZ/7PAzd1+nxKnf48ypflRuD/AP+Zsj0+Qjm9cAX1YmOfulex+W6QRynn2w/qzD8UuLJuNw9S7pbpe2GzT9v71bHdxOa7brrXTaa8DtlyH96l1vQQ8ECdtivlYObu2t4a4O0TbL/r2Lwf7kb5tbi+1no9m+8G+bnah4fqGFxJ+TX0WuCKnjZfTjlAexj48ETjV9t4EHhFZ/lPAx8flA3DeIzfSSFpAvUe9W9RLvpN69fRqETEKkqQnTPTtbQkyh+I3ZaZH530zduRHfUPRSRpkNWUX9FNMawlzSqZecFM17A1PA0iSQ3Ybu4GkSQNNpLTIAsWLMglS5aMomlJ2iHdfPPND2TmwkHzRxLWS5YsYWxsbBRNS9IOKSK+PdF8T4NIUgMMa0lqgGEtSQ0wrCWpAYa1JDXAsJakBhjWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQGGtSQ1wLCWpAYY1pLUAMNakhpgWEtSAwxrSWqAYS1JDTCsJakBhrUkNcCwlqQGTDmsI2JORPy/iLhilAVJkp5uOkfWZwBrRlWIJGmwKYV1RBwALAf+arTlSJL6meqR9Z8CZwFPjbAWSdIAk4Z1RLwW2JCZN0/yvtMjYiwixu6///6hFShJmtqR9SuAX4iIdcBlwHER8aneN2XmBZm5NDOXLly4cMhlStLsNmlYZ+a7M/OAzFwCvB7458x848grkyT9hPdZS1ID5k7nzZl5HXDdSCqRJA3kkbUkNcCwlqQGGNaS1ADDWpIaYFhLUgMMa0lqgGEtSQ0wrCWpAYa1JDXAsJakBhjWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQGGtSQ1wLCWpAYY1pLUAMNakhpgWEtSAwxrSWqAYS1JDTCsJakBhrUkNcCwlqQGGNaS1ADDWpIaYFhLUgMMa0lqgGEtSQ0wrCWpAYa1JDXAsJakBhjWktSAuaNo9Nb1m1iy4spJ37du5fJRfLwk7XA8spakBhjWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQGGtSQ1wLCWpAYY1pLUgEnDOiJ2jYgvR8QtEXF7RPzetihMkrTZVP7bID8CjsvMRyNiZ+CGiLgqM28ccW2SpGrSsM7MBB6tL3eujxxlUZKkLU3pnHVEzImI1cAG4JrMvKnPe06PiLGIGHvysU3DrlOSZrUphXVmPpmZRwIHAEdFxBF93nNBZi7NzKVzdp8/7DolaVab1t0gmbkRuBY4cTTlSJL6mcrdIAsj4tn1+W7Aq4BvjLowSdJmU7kbZD/gooiYQwn3yzPzitGWJUnqmsrdIF8DXrINapEkDeBfMEpSAwxrSWqAYS1JDTCsJakBhrUkNcCwlqQGGNaS1ICp/FHMtL1o0XzGVi4fRdOSNCt5ZC1JDTCsJakBhrUkNcCwlqQGGNaS1ADDWpIaYFhLUgMMa0lqgGEtSQ0wrCWpAYa1JDXAsJakBhjWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQGGtSQ1wLCWpAYY1pLUAMNakhpgWEtSAwxrSWqAYS1JDTCsJakBhrUkNWDuKBq9df0mlqy4chRNb/fWrVw+0yVI2gF5ZC1JDTCsJakBhrUkNcCwlqQGGNaS1ADDWpIaYFhLUgMMa0lqgGEtSQ2YNKwj4sCIuDYivh4Rt0fEGduiMEnSZlP5c/MngDMz86sRsSdwc0Rck5lfH3FtkqRq0iPrzLw3M79anz8CrAEWjbowSdJm0zpnHRFLgJcAN/WZd3pEjEXE2JOPbRpOdZIkYBphHRHzgM8A78jMH/TOz8wLMnNpZi6ds/v8YdYoSbPelMI6InamBPUlmfnZ0ZYkSeo1lbtBAvgEsCYzzx99SZKkXlM5sn4FcDJwXESsro9lI65LktQx6a17mXkDENugFknSAP4FoyQ1wLCWpAYY1pLUAMNakhpgWEtSAwxrSWqAYS1JDTCsJakBU/nvWU/bixbNZ2zl8lE0LUmzkkfWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQGGtSQ1wLCWpAYY1pLUAMNakhpgWEtSAwxrSWqAYS1JDTCsJakBhrUkNcCwlqQGGNaS1ADDWpIaYFhLUgMMa0lqgGEtSQ0wrCWpAYa1JDXAsJakBhjWktQAw1qSGjB3FI3eun4TS1ZcOYqmJWm7tG7l8pG275G1JDXAsJakBhjWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQGGtSQ1YNKwjogLI2JDRNy2LQqSJD3dVI6sVwEnjrgOSdIEJg3rzLweeGgb1CJJGsBz1pLUgKGFdUScHhFjETH25GObhtWsJIkhhnVmXpCZSzNz6Zzd5w+rWUkSngaRpCZM5da9S4F/Aw6NiO9GxGmjL0uS1DXp/ykmM9+wLQqRJA3maRBJaoBhLUkNMKwlqQGGtSQ1wLCWpAYY1pLUAMNakhpgWEtSAyb9o5it8aJF8xlbuXwUTUvSrOSRtSQ1wLCWpAYY1pLUAMNakhpgWEtSAwxrSWqAYS1JDTCsJakBhrUkNcCwlqQGGNaS1ADDWpIaYFhLUgMiM4ffaMQjwB1Db7gdC4AHZrqIGTbbx2C29x8cA5jeGCzOzIWDZo7kP5EK3JGZS0fU9nYvIsZmc//BMZjt/QfHAIY7Bp4GkaQGGNaS1IBRhfUFI2q3FbO9/+AYzPb+g2MAQxyDkVxglCQNl6dBJKkBhrUkNWCoYR0RJ0bEHRGxNiJWDLPt7U1ErIuIWyNidUSM1Wl7R8Q1EXFn/XevOj0i4sN1XL4WES+d2eqnLyIujIgNEXFbZ9q0+xsRp9T33xkRp8xEX7bWgDE4NyLW1+1gdUQs68x7dx2DOyLihM70JveTiDgwIq6NiK9HxO0RcUadPmu2gwnGYPTbQWYO5QHMAe4CDgZ2AW4BDh9W+9vbA1gHLOiZ9gFgRX2+AvjD+nwZcBUQwMuAm2a6/q3o7zHAS4Hbtra/wN7A3fXfverzvWa6b89wDM4F3tnnvYfXfeBZwEF135jT8n4C7Ae8tD7fE/hm7ees2Q4mGIORbwfDPLI+ClibmXdn5n8AlwEnDbH9FpwEXFSfXwT8Ymf6xVncCDw7IvabiQK3VmZeDzzUM3m6/T0BuCYzH8rMh4FrgBNHX/1wDBiDQU4CLsvMH2Xmt4C1lH2k2f0kM+/NzK/W548Aa4BFzKLtYIIxGGRo28Eww3oRcE/n9XeZuBOtS+AfI+LmiDi9Tts3M++tz78P7Fuf76hjM93+7qjj8N/rz/wLx08BsIOPQUQsAV4C3MQs3Q56xgBGvB14gXHrHZ2ZLwVeA7wtIo7pzszyG2jW3Bc52/rb8RfAIcCRwL3AB2e2nNGLiHnAZ4B3ZOYPuvNmy3bQZwxGvh0MM6zXAwd2Xh9Qp+2QMnN9/XcD8DnKz5r7xk9v1H831LfvqGMz3f7ucOOQmfdl5pOZ+RTwccp2ADvoGETEzpSQuiQzP1snz6rtoN8YbIvtYJhh/RXgeRFxUETsArwe+PwQ299uRMQeEbHn+HPg1cBtlP6OX9k+Bfi7+vzzwJvq1fGXAZs6PxtbNt3+Xg28OiL2qj8TX12nNavn2sMvUbYDKGPw+oh4VkQcBDwP+DIN7ycREcAngDWZeX5n1qzZDgaNwTbZDoZ8pXQZ5eroXcDZM33ldlQPyhXcW+rj9vG+AvsAXwTuBP4J2LtOD+AjdVxuBZbOdB+2os+XUn7e/Zhyfu20rekv8BbKRZa1wJtnul9DGINP1j5+re5s+3Xef3YdgzuA13SmN7mfAEdTTnF8DVhdH8tm03YwwRiMfDvwz80lqQFeYJSkBhjWktQAw1qSGmBYS1IDDGtJaoBhLUkNMKwlqQH/HxzCT5yWChwZAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.aspects.apply(lambda x: len(x)).value_counts().plot(kind='barh', title='Distribution of Number of Aspects per sentence/text');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Aspects combinations distribution per text/sentence"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We further look at all possible combinations of aspects per text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x1008 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(14,14))\n",
    "df.aspects.apply(lambda x:[item[0] for item in x]).value_counts().plot(kind='barh', title='Distribution of Combination of aspects per sentence/text');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Number of sentences per text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def count_sentences(text):\n",
    "    return len(list(nlp(text).sents))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "count_sentences(\"I have two sentences here. spaCy must correctly detect the number of sentences.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "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": [
    "df.text.apply(count_sentences).value_counts().plot(kind='barh', title='Distribution of Number of sentences per text');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preprocess Texts"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Coreference Resolution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "import neuralcoref"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<spacy.lang.en.English at 0x7fb6732b46a0>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neuralcoref.add_to_pipe(nlp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def replace_pronouns(text):\n",
    "    \"\"\"Replace pronouns in sentences with co-refering nouns\"\"\"\n",
    "    doc = nlp(text)\n",
    "    return doc._.coref_resolved"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Anz2 is developer. Anz2 is coding right now :).'"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "replace_pronouns(\"Anz2 is developer. He is coding right now :).\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.loc[:, 'text'] = df.text.apply(replace_pronouns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Extract Aspect Term Candidates from sentences"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We just try to extract aspect term candidates from given text. We take only nouns which are longer than 3 characters.\n",
    "That candidates aren't used in our approach but just to see and check candidate terms. Similar approach is used for extracting opinion words and referring aspect terms later."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "def extract_candidate_terms(text):\n",
    "    \"\"\"Extracts all nouns from text as aspect term candidates\"\"\"\n",
    "    \n",
    "    req_tag = ['NN']\n",
    "    candidate_terms = []\n",
    "    \n",
    "    for token in nlp(text):\n",
    "        if (\n",
    "            token.tag_ in req_tag and # token must be noun\n",
    "            token.shape_ != 'x' and # token mustn't be only one character\n",
    "            token.shape_ != 'xx' and # token mustn't be only two character\n",
    "            token.shape_ != 'xxx' # token mustn't be only three character\n",
    "        ):\n",
    "            candidate_terms.append(token.lemma_)\n",
    "            \n",
    "    return candidate_terms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('But the staff was so horrible to us.', ['staff'])"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[0], extract_candidate_terms(df.text.iloc[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['aspect_term_candidates'] = df.text.apply(extract_candidate_terms)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Aspect Term Classification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We could classify aspect term candidates using word embedding similarity. We are taking embedding of aspect term candidate, also embeddings of aspects and find maximum match. If it's higher than predefined threshold (hyperparameter which can be tuned in final end-to-end system) we assign given aspect category."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>text</th>\n",
       "      <th>terms</th>\n",
       "      <th>aspects</th>\n",
       "      <th>aspect_term_candidates</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>But the staff was so horrible to us.</td>\n",
       "      <td>[(staff, negative)]</td>\n",
       "      <td>[(service, negative)]</td>\n",
       "      <td>[staff]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>To be completely fair, the only redeeming fact...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[(food, positive), (anecdotes/miscellaneous, n...</td>\n",
       "      <td>[factor, food]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>The food is uniformly exceptional, with a very...</td>\n",
       "      <td>[(food, positive), (kitchen, positive), (menu,...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[food, kitchen, food, menu]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Where Gabriela personaly greets you and recomm...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(service, positive)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>For those that go once and don't enjoy it, all...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(anecdotes/miscellaneous, positive)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  \\\n",
       "0               But the staff was so horrible to us.   \n",
       "1  To be completely fair, the only redeeming fact...   \n",
       "2  The food is uniformly exceptional, with a very...   \n",
       "3  Where Gabriela personaly greets you and recomm...   \n",
       "4  For those that go once and don't enjoy it, all...   \n",
       "\n",
       "                                               terms  \\\n",
       "0                                [(staff, negative)]   \n",
       "1                                 [(food, positive)]   \n",
       "2  [(food, positive), (kitchen, positive), (menu,...   \n",
       "3                                                 []   \n",
       "4                                                 []   \n",
       "\n",
       "                                             aspects  \\\n",
       "0                              [(service, negative)]   \n",
       "1  [(food, positive), (anecdotes/miscellaneous, n...   \n",
       "2                                 [(food, positive)]   \n",
       "3                              [(service, positive)]   \n",
       "4              [(anecdotes/miscellaneous, positive)]   \n",
       "\n",
       "        aspect_term_candidates  \n",
       "0                      [staff]  \n",
       "1               [factor, food]  \n",
       "2  [food, kitchen, food, menu]  \n",
       "3                           []  \n",
       "4                           []  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "class EmbeddingClassifier:\n",
    "    \"\"\"Classifying aspect terms into aspects using word embedding similarity\"\"\"\n",
    "    \n",
    "    def __init__(self, nlp, classes:list, threshold:int=0.4):\n",
    "        self.classes = classes\n",
    "        self.class_spans = [nlp(item)[:] for item in classes]\n",
    "        self.threshold = threshold\n",
    "        self.nlp = nlp\n",
    "    \n",
    "    def predict(self, items:list):\n",
    "        \"\"\"Match items to classes using embeddings\"\"\"\n",
    "        predicted_classes = set()\n",
    "        \n",
    "        for item in items:\n",
    "            class_similarities = [\n",
    "                nlp(item)[:].similarity(class_item) \n",
    "                for class_item in self.class_spans\n",
    "            ]\n",
    "            max_idx = np.argmax(class_similarities)\n",
    "            if class_similarities[max_idx] > self.threshold:\n",
    "                predicted_classes.add(self.classes[max_idx])\n",
    "        \n",
    "        return list(predicted_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'ambience', 'anecdotes/miscellaneous', 'food', 'price', 'service'}"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "set(df.aspects.apply(lambda x: [item[0] for item in x]).sum())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`anecdotes/miscellaneous` is bad aspect category name and will be ignored in our evaluation measurements. It's more general and can't be easily used for measuring vector similarity. To solve such issues we could take several center words for each aspect category like: **food** - **[pizza, drinks, soup, hamburger, fries, etc]** and then measure maximum match between them to decide similarity of aspect term candidate to given aspect category."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "aspect_classes = ['ambience', 'miscellaneous', 'food', 'price', 'service']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf = EmbeddingClassifier(nlp=nlp, classes=aspect_classes, threshold=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>text</th>\n",
       "      <th>terms</th>\n",
       "      <th>aspects</th>\n",
       "      <th>aspect_term_candidates</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>But the staff was so horrible to us.</td>\n",
       "      <td>[(staff, negative)]</td>\n",
       "      <td>[(service, negative)]</td>\n",
       "      <td>[staff]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>To be completely fair, the only redeeming fact...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[(food, positive), (anecdotes/miscellaneous, n...</td>\n",
       "      <td>[factor, food]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>The food is uniformly exceptional, with a very...</td>\n",
       "      <td>[(food, positive), (kitchen, positive), (menu,...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[food, kitchen, food, menu]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Where Gabriela personaly greets you and recomm...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(service, positive)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>For those that go once and don't enjoy it, all...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(anecdotes/miscellaneous, positive)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  \\\n",
       "0               But the staff was so horrible to us.   \n",
       "1  To be completely fair, the only redeeming fact...   \n",
       "2  The food is uniformly exceptional, with a very...   \n",
       "3  Where Gabriela personaly greets you and recomm...   \n",
       "4  For those that go once and don't enjoy it, all...   \n",
       "\n",
       "                                               terms  \\\n",
       "0                                [(staff, negative)]   \n",
       "1                                 [(food, positive)]   \n",
       "2  [(food, positive), (kitchen, positive), (menu,...   \n",
       "3                                                 []   \n",
       "4                                                 []   \n",
       "\n",
       "                                             aspects  \\\n",
       "0                              [(service, negative)]   \n",
       "1  [(food, positive), (anecdotes/miscellaneous, n...   \n",
       "2                                 [(food, positive)]   \n",
       "3                              [(service, positive)]   \n",
       "4              [(anecdotes/miscellaneous, positive)]   \n",
       "\n",
       "        aspect_term_candidates  \n",
       "0                      [staff]  \n",
       "1               [factor, food]  \n",
       "2  [food, kitchen, food, menu]  \n",
       "3                           []  \n",
       "4                           []  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['service']"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.predict(df.loc[0, 'aspect_term_candidates'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['food']"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.predict(df.loc[1, 'aspect_term_candidates'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['food']"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.predict(df.loc[2, 'aspect_term_candidates'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.predict(df.loc[3, 'aspect_term_candidates'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Detecting Positive and Negative opinion words in context  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This approach just checks words in opinion words list, so no similarity is used here. We find negative or positive opinion words and then try to find modifiers of given opinion like: **very** `bad` service. Here very is increasing negativeness of `bad` and it refers to service. We store that negative sentiment referring to service and then try to assign aspect category to service using previously explained similarity based approach."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get negative and positive English opinion words data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists('absa_data/neg_words.txt'):\n",
    "    !cd absa_data/ && wget https://raw.githubusercontent.com/LFCapollo/Aspect-Based-Sentiment-Analysis/master/data/opinion-lexicon-English/neg_words.txt\n",
    "if not os.path.exists('absa_data/pos_words.txt'):\n",
    "    !cd absa_data/ && wget https://raw.githubusercontent.com/LFCapollo/Aspect-Based-Sentiment-Analysis/master/data/opinion-lexicon-English/pos_words.txt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Parse and Load opinion words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_opinions(path:str):\n",
    "    \"\"\"Load opinion words from file\"\"\"\n",
    "    opinions = []\n",
    "    \n",
    "    with open(path, 'r',  encoding=\"ISO-8859-1\") as file:\n",
    "        for line in file.readlines():\n",
    "            if line.strip():\n",
    "                opinions.append(line.lower().strip())\n",
    "    \n",
    "    \n",
    "    return set(opinions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "negative_words = load_opinions('absa_data/neg_words.txt')\n",
    "positive_words = load_opinions('absa_data/pos_words.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['passive', 'addicted', 'burn', 'thug', 'conspiratorial']"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(negative_words)[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['fav', 'fervently', 'applaud', 'sparkle', 'delicacy']"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(positive_words)[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Scoring of feature sentiments using dependency parsing "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "HINT: Please use [spacy dependency visualizer](https://explosion.ai/demos/displacy?text=&model=en_core_web_sm&cpu=1&cph=1) to understand examples mentioned here "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "def feature_sentiment(sentence:str, pos:set, neg:set):\n",
    "    \"\"\"Analyses sentiment of each aspect term in sentence using negative and positive opinion words list\"\"\"\n",
    "    \n",
    "    sent_dict = dict()\n",
    "    sentence = nlp(sentence)\n",
    "    \n",
    "    for token in sentence:\n",
    "        # check if the word is an opinion word, then assign sentiment\n",
    "        if token.text in neg or token.text in pos:\n",
    "            sentiment = 1 if token.text in pos else -1\n",
    "            \n",
    "            # [1] if target is an adverb modifier (i.e. pretty, highly, etc.)\n",
    "            # but happens to be an opinion word, ignore and pass\n",
    "            # example: Client eats [amazingly] rotten fruit.\n",
    "            if (token.dep_ == \"advmod\"):\n",
    "                continue\n",
    "            \n",
    "            # [2] if target is an adjectival modifier (i.e. bad, great, etc.) \n",
    "            # it directly refers to given noun and must assign sentiment of given adjective to given \n",
    "            # token head which is modified noun\n",
    "            # example: Client eats a [rotten] fruit.\n",
    "            elif (token.dep_ == \"amod\"):\n",
    "                sent_dict[token.head.text] = sentiment\n",
    "                \n",
    "            # for opinion words that are adjectives, adverbs, verbs...\n",
    "            else:\n",
    "                for child in token.children:\n",
    "                    \n",
    "                    # [3] if there's a adj modifier (i.e. very, pretty, etc.) add more weight to sentiment\n",
    "                    # This could be better updated for modifiers that either positively or negatively emphasize\n",
    "                    # example: I[like] their service very much.\n",
    "                    if (\n",
    "                        (child.dep_ == \"amod\") or (child.dep_ == \"advmod\")\n",
    "                    ) and (\n",
    "                        child.text in neg or child.text in pos\n",
    "                    ):\n",
    "                        sentiment *= 1.5\n",
    "                        \n",
    "                    # [4] check for negation words and flip the sign of sentiment\n",
    "                    # example: I don't [like] their service that much.\n",
    "                    if child.dep_ == \"neg\":\n",
    "                        sentiment *= -1\n",
    "                \n",
    "                # [5] in case of VERB we check directly referred objects to directly assign sentiment\n",
    "                # example: I [like] their service.\n",
    "                for child in token.children:\n",
    "                    \n",
    "                    # if verb, check if there's a direct object\n",
    "                    if (token.pos_ == \"VERB\") & (child.dep_ == \"dobj\"):                        \n",
    "                        sent_dict[child.text] = sentiment\n",
    "                        \n",
    "                        # check for conjugates (a AND b), then add both to dictionary\n",
    "                        # example: I [like] their service and food.\n",
    "                        subchildren = []\n",
    "                        is_conj = False\n",
    "                        for subchild in child.children:\n",
    "                            if subchild.text == \"and\":\n",
    "                                is_conj = True\n",
    "                            if is_conj and subchild.text != \"and\":\n",
    "                                subchildren.append(subchild.text)\n",
    "                                is_conj = False\n",
    "                                \n",
    "                        for subchild in subchildren:\n",
    "                            sent_dict[subchild] = sentiment\n",
    "\n",
    "                # check another childrens of parent in dependency tree \n",
    "                for child in token.head.children:\n",
    "                    \n",
    "                    if (\n",
    "                        (child.dep_ == \"amod\") or (child.dep_ == \"advmod\")\n",
    "                    ) and (\n",
    "                        child.text in neg or child.text in pos\n",
    "                    ):\n",
    "                        sentiment *= 1.5\n",
    "                        \n",
    "                    # check for negation words and flip the sign of sentiment\n",
    "                    if (child.dep_ == \"neg\"): \n",
    "                        sentiment *= -1\n",
    "                \n",
    "                # check for nouns and assign sentiment to it (or to compound noun)\n",
    "                for child in token.head.children:\n",
    "                    \n",
    "                    noun = \"\"\n",
    "                    if (child.pos_ == \"NOUN\") and (child.text not in sent_dict):\n",
    "                        noun = child.text\n",
    "                        \n",
    "                        # [7] Check for compound nouns\n",
    "                        # example: Their entrance door is amazing and beautiful.\n",
    "                        for subchild in child.children:\n",
    "                            if subchild.dep_ == \"compound\":\n",
    "                                noun = subchild.text + \" \" + noun\n",
    "                        sent_dict[noun] = sentiment\n",
    "                    \n",
    "    return sent_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0                        {'staff': -1}\n",
       "1    {'factor': 1, 'deficiencies': -1}\n",
       "2            {'food': 1, 'kitchen': 1}\n",
       "3                                   {}\n",
       "4                           {'it': -1}\n",
       "Name: text, dtype: object"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.head().apply(lambda x: feature_sentiment(sentence=x, pos=positive_words, neg=negative_words))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Aspect Based Sentiment Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Implementation of aspect based sentiment analysis class object. It takes text, split into sentences, find aspect term candidates or nouns in sentences and referred sentiments with previously described scoring mechanism. It then assigns aspect categories for each aspect term candidate."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "class AspectBasedSentimentAnalyser:\n",
    "    \"\"\"Implements aspect based sentiment analysis\"\"\"\n",
    "    \n",
    "    def __init__(self, nlp, aspects:list, negative_opinions:list, positive_opinions:list, threshold:int):\n",
    "        self.classifier = EmbeddingClassifier(nlp=nlp, classes=aspects, threshold=threshold)\n",
    "        self.negative_opinions = negative_opinions\n",
    "        self.positive_opinions = positive_opinions\n",
    "        self.nlp = nlp\n",
    "    \n",
    "    def analyse(self, text):\n",
    "        aspect_sentiments = {}\n",
    "        for sentence in self.nlp(text).sents:\n",
    "            for feature, sentiment_value in feature_sentiment(\n",
    "                sentence=sentence.text, pos=self.positive_opinions, neg=self.negative_opinions\n",
    "            ).items():\n",
    "                predicted_aspect = self.classifier.predict([feature])\n",
    "                \n",
    "                if predicted_aspect:\n",
    "                    aspect_sentiments[predicted_aspect[0]] = aspect_sentiments.get(predicted_aspect[0], 0) + sentiment_value\n",
    "        \n",
    "        return aspect_sentiments"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "sent_analyser = AspectBasedSentimentAnalyser(\n",
    "    nlp=nlp,\n",
    "    aspects=aspect_classes,\n",
    "    negative_opinions=negative_words,\n",
    "    positive_opinions=positive_words,\n",
    "    threshold=0.3\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>text</th>\n",
       "      <th>terms</th>\n",
       "      <th>aspects</th>\n",
       "      <th>aspect_term_candidates</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>But the staff was so horrible to us.</td>\n",
       "      <td>[(staff, negative)]</td>\n",
       "      <td>[(service, negative)]</td>\n",
       "      <td>[staff]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>To be completely fair, the only redeeming fact...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[(food, positive), (anecdotes/miscellaneous, n...</td>\n",
       "      <td>[factor, food]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>The food is uniformly exceptional, with a very...</td>\n",
       "      <td>[(food, positive), (kitchen, positive), (menu,...</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[food, kitchen, food, menu]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Where Gabriela personaly greets you and recomm...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(service, positive)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>For those that go once and don't enjoy it, all...</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(anecdotes/miscellaneous, positive)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Not only was the food outstanding, but the lit...</td>\n",
       "      <td>[(food, positive), (perks, positive)]</td>\n",
       "      <td>[(food, positive), (service, positive)]</td>\n",
       "      <td>[food]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>It is very overpriced and not very tasty.</td>\n",
       "      <td>[]</td>\n",
       "      <td>[(food, negative), (price, negative)]</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Our agreed favorite is the orrechiete with sau...</td>\n",
       "      <td>[(orrechiete with sausage and chicken, positiv...</td>\n",
       "      <td>[(food, positive), (service, positive)]</td>\n",
       "      <td>[favorite, orrechiete, sausage, chicken, dish,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>The Bagels have an outstanding taste with a te...</td>\n",
       "      <td>[(Bagels, positive)]</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[taste, texture]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Nevertheless the food itself is pretty good.</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[(food, positive)]</td>\n",
       "      <td>[food]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  \\\n",
       "0               But the staff was so horrible to us.   \n",
       "1  To be completely fair, the only redeeming fact...   \n",
       "2  The food is uniformly exceptional, with a very...   \n",
       "3  Where Gabriela personaly greets you and recomm...   \n",
       "4  For those that go once and don't enjoy it, all...   \n",
       "5  Not only was the food outstanding, but the lit...   \n",
       "6          It is very overpriced and not very tasty.   \n",
       "7  Our agreed favorite is the orrechiete with sau...   \n",
       "8  The Bagels have an outstanding taste with a te...   \n",
       "9       Nevertheless the food itself is pretty good.   \n",
       "\n",
       "                                               terms  \\\n",
       "0                                [(staff, negative)]   \n",
       "1                                 [(food, positive)]   \n",
       "2  [(food, positive), (kitchen, positive), (menu,...   \n",
       "3                                                 []   \n",
       "4                                                 []   \n",
       "5              [(food, positive), (perks, positive)]   \n",
       "6                                                 []   \n",
       "7  [(orrechiete with sausage and chicken, positiv...   \n",
       "8                               [(Bagels, positive)]   \n",
       "9                                 [(food, positive)]   \n",
       "\n",
       "                                             aspects  \\\n",
       "0                              [(service, negative)]   \n",
       "1  [(food, positive), (anecdotes/miscellaneous, n...   \n",
       "2                                 [(food, positive)]   \n",
       "3                              [(service, positive)]   \n",
       "4              [(anecdotes/miscellaneous, positive)]   \n",
       "5            [(food, positive), (service, positive)]   \n",
       "6              [(food, negative), (price, negative)]   \n",
       "7            [(food, positive), (service, positive)]   \n",
       "8                                 [(food, positive)]   \n",
       "9                                 [(food, positive)]   \n",
       "\n",
       "                              aspect_term_candidates  \n",
       "0                                            [staff]  \n",
       "1                                     [factor, food]  \n",
       "2                        [food, kitchen, food, menu]  \n",
       "3                                                 []  \n",
       "4                                                 []  \n",
       "5                                             [food]  \n",
       "6                                                 []  \n",
       "7  [favorite, orrechiete, sausage, chicken, dish,...  \n",
       "8                                   [taste, texture]  \n",
       "9                                             [food]  "
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('But the staff was so horrible to us.', {'service': -1})"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[0], sent_analyser.analyse(df.text.iloc[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(\"To be completely fair, the only redeeming factor was the food, which was above average, but couldn't make up for all the other deficiencies of Teodora.\",\n",
       " {'price': 1})"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[1], sent_analyser.analyse(df.text.iloc[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(\"The food is uniformly exceptional, with a very capable kitchen which will proudly whip up whatever you feel like eating, whether The food's on the menu or not.\",\n",
       " {'food': 2})"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[2], sent_analyser.analyse(df.text.iloc[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('Where Gabriela personaly greets you and recommends you what to eat.', {})"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[3], sent_analyser.analyse(df.text.iloc[3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(\"For those that go once and don't enjoy it, all I can say is that those that go once and don't enjoy it just don't get it.\",\n",
       " {'food': -1})"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[4], sent_analyser.analyse(df.text.iloc[4])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(\"Not only was the food outstanding, but the little 'perks' were great.\",\n",
       " {'food': 1})"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[5], sent_analyser.analyse(df.text.iloc[5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('Nevertheless the food itself is pretty good.', {'food': 1.5})"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.iloc[9], sent_analyser.analyse(df.text.iloc[9])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluation of results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We simply calculate confusion matrix values with custom functions to consider specific cases."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "def convert_prediction(prediction):\n",
    "    return [\n",
    "        (\n",
    "            item if item!='miscellaneous' else 'anecdotes/miscellaneous', \"positive\" if value > 0 else 'neutral' if value==0 else \"negative\"\n",
    "        ) for item, value in prediction.items()\n",
    "    ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('anecdotes/miscellaneous', 'negative')]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "convert_prediction({'miscellaneous': -1})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('food', 'neutral')]"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "convert_prediction({'food': 0})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def confusion_value(label, prediction):\n",
    "    result = None\n",
    "    if not label:\n",
    "        if not prediction:\n",
    "            result = 'TN'\n",
    "        else:\n",
    "            result = 'FP'\n",
    "    else:\n",
    "        if not prediction:\n",
    "            result = 'FN'\n",
    "        else:\n",
    "            converted_prediction = convert_prediction(prediction)\n",
    "            result = 'TP'\n",
    "            for item in label:\n",
    "                if item in converted_prediction or item[0] == 'anecdotes/miscellaneous':\n",
    "                    pass\n",
    "                else:\n",
    "                    result = 'FN'\n",
    "                    break\n",
    "        \n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "def confusion_matrix_values(labels, predictions):\n",
    "    confusion_matrix = {}\n",
    "    \n",
    "    for label, prediction in zip(labels, predictions):\n",
    "        conf_value = confusion_value(label, prediction)\n",
    "        confusion_matrix[conf_value] = confusion_matrix.get(conf_value, 0) + 1\n",
    "    \n",
    "    return confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0                                [(service, negative)]\n",
       "1    [(food, positive), (anecdotes/miscellaneous, n...\n",
       "2                                   [(food, positive)]\n",
       "3                                [(service, positive)]\n",
       "4                [(anecdotes/miscellaneous, positive)]\n",
       "Name: aspects, dtype: object"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.aspects.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    {'service': -1}\n",
       "1       {'price': 1}\n",
       "2        {'food': 2}\n",
       "3                 {}\n",
       "4       {'food': -1}\n",
       "dtype: object"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.Series(sent_analyser.analyse(text) for text in df.text.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "df.loc[:, 'aspect_predictions'] = df.text.apply(sent_analyser.analyse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "conf_matrix = confusion_matrix_values(df.aspects.values.tolist(), df.aspect_predictions.values.tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'TP': 817, 'FN': 2227}"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "conf_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we see we don't have false positive cases and true negative cases. That can be simply explained. False positive isn't available because we are using fixed set of opinion words which keeps us from using false opinion words. Also we are lacking of text samples without classified sentiments.\n",
    "\n",
    "Accuracy of our system is quite low but we have 100% precision and just recall is low. It can be improved with more advanced approaches. Here we used simplest possible approach which is easily adaptable to any case and what we need to make this work is to list aspect categories."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualize Aspect Based Sentiment Analysis Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Visualisation of aspect sentiments is done with positive, neutral and negative cases only (`contradiction` categories are missing to avoid misconfusions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "def summary_aspect_based_sentiment_analysis(predictions):\n",
    "    summary = {}\n",
    "    for prediction in predictions:\n",
    "        if prediction:\n",
    "            for aspect, sentiment in convert_prediction(prediction):\n",
    "                if aspect not in summary:\n",
    "                    summary[aspect] = {}\n",
    "\n",
    "                summary[aspect][sentiment] = summary[aspect].get(sentiment, 0) + 1\n",
    "    \n",
    "    return summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "summary = summary_aspect_based_sentiment_analysis(df.aspect_predictions.values.tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'service': {'negative': 95, 'positive': 298, 'neutral': 3},\n",
       " 'price': {'positive': 105, 'negative': 36, 'neutral': 1},\n",
       " 'food': {'positive': 817, 'negative': 188, 'neutral': 38},\n",
       " 'ambience': {'positive': 145, 'negative': 28, 'neutral': 2},\n",
       " 'anecdotes/miscellaneous': {'negative': 2, 'positive': 4}}"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 1008x1008 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "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.figure(figsize=(14,14));\n",
    "pd.DataFrame(summary).T.plot(kind='bar', title='Aspect Based Sentiment Analysis');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# References"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[Nika Kvedelidze Sample Project at MaxinAI](https://github.com/LFCapollo/Aspect-Based-Sentiment-Analysis)\n",
    "\n",
    "[Intellica.AI - ABSA blog](https://medium.com/@Intellica.AI/aspect-based-sentiment-analysis-everything-you-wanted-to-know-1be41572e238)\n",
    "\n",
    "[remicnrd - ABSA blog](https://remicnrd.github.io/Aspect-based-sentiment-analysis) - ideas"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.0"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}