Skip to content

Latest commit

 

History

History
235 lines (195 loc) · 11.8 KB

TODO.md

File metadata and controls

235 lines (195 loc) · 11.8 KB

TODO List

In Progress

  • [ ]

Todo

  • Fix music bot ---> player only in one channel for now pls

  • FIX music bot ---> private videos break it

  • AUTO TIMED VOCAB PRACTICE !!! DONE --> voice recognition next!

  • surveillance.py: Voice stats graphs, Process old logs

  • Optimization of everything:

    • Pylint (errs, with cmd! -> 9.0 score above)
    • Async
    • Logs (function calls?!)
    • Document/Test (ai_algo.py - _apply_actions test (rmv))
    • ChatGPT on Discord
    • ChatGPT Base forms of vocab occurences in listening/reading files
    • Discord Activity + Music monitoring (Voice, Music, Korean)

  • music.py:

    • UPD: Deep opt., polishing with other bots as insp., apply logs instead of prints (+wrappers?)
    • IMP: Most played dropdown choose at all times; auto select with distribution (radio)

  • korean.py

    • UPD format the output more nicely after session is done (check opts; clarify notes for certain session types with bold font)

    • UPD inform that you can uncover the expressions to see the answer (help cmd)

    • IMP Notes command? /help only for the session, too much text might drive users away

    • UPD Note - more meanings chaos (with numbers, without numbers[;;;])

    • IMP Click to start button (resolve pressing the same option in dropdown) ---> FIX: ending the session and fast opening a new one will crash (cuz saving worksheets!)

    • FIX/IMP: Left voice -> bot left, and forgot about session in text channel (no guesses(?)), automatically cancel the session upon leave; Remind to leave the voice chat when its over, or disconnect the user (one more button!) / Auto connect + Auto disconnect from voice when interacting with interface

    • PORTFOLIO: Present on website - Show the graphs of right/wrongs for each person

    • IMP: Speech recognition (not only google's voice.. [because of phrases later!])

    • IMP: Audio downloading (current script is published in memo bookmark)

    • IMP: Auto-repeat (check all audio get max length) + stop sound if buttoned

    • IMP: New selection (no selection needed, auto calculate the best selection - new words (no 50!) / review)

    • IMP: Add Grammar

    • IMP: Vocab/User data opt. for publishing (Elastic search instead of gsheet)

    • IMP: Translation (for reading/listening)

    • IMP: Session visualization (streak play, github-like)

    • UPD: CHECK Rearranged columns in google sheets --> might need update for that, check it

    • ???: CHECK Commit - [UPD] move connect/disconnect from music.py into shared utils https://github.com/freezpmark/personal-discord-bot/commit/e69a415341308cc6004442a331903421b5692099#diff-35606ac1b506923433a05188f57dfd000f0b13a62abc984a0ea8f41ab0a14fbf

    • LLM/ChatGPT IMP:

      • IMP: Questions (reading & listening)
      • IMP: Create expressions (for listening/reading, grammar/vocab)
      • IMP: Spreadsheet update with extra info (issue with Example column - synonyms not recognized)
      • IMP: Stable Diffusion - add more pics (automate without checking them) [text -> img -> text -> same text? (supervised)]
      • IMP: LLM specialized for base form detection (for Reading/Listening to build vocab)
      • IMP: Gspread: Phrases/Two words - split, find words, if both exist, create () with 를/을 ...
      • IMP: Similarity between words in vocab session (meanings/spelling/antonyms)
      • IMP: Buttons for Listening, Reading (exploration with -> grammar used, vocab used, translation)

    • IDEAS TO IMPLEMENT:

      • Take into account on which days u guessed it.. if u still know after long time -> alot of points!, For each in the same time, less points ---- prove with data science!... less chance to pick something from the same day..! Already there but.. more sophisticated
      • Writing vocab f.: vocab_writing [REM] vocab writing function
      • Audio for reading too (?) Lvl 3 already done

    • LLM IDEAS TO IMPLEMENT:

      • D&D idea of learning (inventory - nouns, adjs, etc. --- together/solo)
      • Create your own learning vocab list! (getting the vocab - kdrama, sub.)
      • Creating content with your Vocab
      • Clustering words relations (similar words opts!)
      • Well known story concept .. harry potter... take ur favo char... Be with it
      • Create dynamic adaptable content... sth they can build on, commit
      • Knowledge graph? Similarity between Korean words (spaced in dash app) [2 syllables same, same base english word]
PROMPT EXAMPLES:
Image creating:
 - How would you describe word 당연히 by a picture? Simple, cartoonic image that could be generated by AI Image Generator.

Similarity words prompt:
 - Find similar words that could be hard to understand the meaning difference
 - Detect similar words which are often hard for people to understand the difference in the meaning between them
 - Detect similar words that are hard to distinguish their meaning from each other for people. The ones that people usually search on the internet with a question: "What is the difference between these two words?"
Use these words below:

Done ✓

  • Create my first TODO.md

  • Update Korean module notes:

Per Level (START)

  • get reading and listening text into data files

Per Lesson (START):

  • Google Docs (Grammar):

    • Process Hongik+Topik (add translations, make it more clear)

  • Google Sheets (Vocab):

    • merge Book + Freq + Topik words into one cell
    • expr. like 한국 사람 into separate words (except for using verbs (하다)

  • Use POS Tagger (gsheet4) and manually check listening/writing text files

  • Add sound labels (gsheet5)

  • Discord Learning:

    • learn the vocab
    • practice listening + writing
    • During the learning of vocab, fix ","'s and ";"s but also the words in general (after listening, the meaning might be different)

Per Level (END) Check grammar again with: TTMIK + Grammar Book + Mirinae

POS TAGGER PROCESS - filling Listening_Used / Reading_Used in Sheets [LVL 1-2 sheet only]

  • Play the listening file, or read if it is reading. Fix the mistakes that were made by Google Lens text transition from picture.
  • Turn on POS Tagger (gsheet4)
    1. Check whether we have any duplicates, if yes, we should search the meanings in the text. According to that we might add a word (which we already have, but with different meaning) with a different number
    2. Basic word checking:
    • check numbered words whether the meaning corresponds to the text
    • check whether the base form of the word is correct
    • Out of lesson (in previous) - no need to do anything
    • Out of lesson (in next) - modify Lesson column into current Lesson
    • Fill added, cuz 사! is number, but it was 사요 -> 사다.. check forms!

-- correcting number of the meaning of the word with {number} -- wrongly formed word, add it corrected (++ attachable) ~~ discard - grammar point / wrongly parsed word ++ new word

Update occurence tabs:

  • Iterate over the words again, and what is marked as above, we should modify/delete the written listening/writing occurences tabs
  • As for "All out of vocab" marked words, we have to add them since they were not present in our vocab at all

GOOGLE SPREADSHEET INFO Korea - Vocabulary (raw) Tabs:

  • raw - raw source of vocabulary that was made during my lessons
  • arranged - easier to read version of vocabulary of raw
  • freq unique - words that are not present in my vocab (informational purpose)
  • topi unique - words that are not present in my vocab (informational purpose)
  • freq - all words that are present in freq. book

Korea - Vocabulary Tabs:

  • listening/reading occurences shall be merged together, group rows by Word too
  • Copy of ... tab - serves as backup
  • Dupl. meanings - synonyms that are grammarly used differently, we already have such words included, it is just slightly different on the type of word (mentioned in Abbreviations)
  • Grammar pts. - potential content to use for grammar docs (waiting for being used in writing/listening and then mention it!)
  • Abbreviations - clarification of abbreviations of freq's content

Korea - Users stats

  • each user has a tab where logs of vocab practice is stored

PROGRAM PLAN Korean Program (individuality value added, no flashcards needed!!!) Vocabulary <> Tracks (Play, Pause, Back, Replay, Next, Previous) [display/hide text] Stats for both! Grammar part - later!

Columns: Presence in topik/freq [underline], listening/reading [bold] files, audio [no spoil]

Installing KoNLPy https://konlpy.org/ 3. https://www.youtube.com/watch?v=104dNWmM6Rs 4. Find JPype1‑1.4.0‑cp310‑cp310‑win_amd64.whl, download it and put it into folder from which you will install it via command

  • 310 means 3.10 Python
  • 64 means Windows 64 bit ver.

Installing ffmpeg for voice: https://www.youtube.com/watch?v=IECI72XEox0

Guide?: python -m venv venv call venv/Scripts/activate.bat pip install -e .

Korean Study Guide Text text_general = """ /vocab (vocabulary learning) /listen (listening practice) /read (reading practice)

Followed by the command, you have to type a certain number that determines what kind of lesson is going to be picked. There are 4 levels and each contains 30 lessons. ㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡ """ text_vocab = """

  1. One (1) - finds and starts the next user's unknown lesson session
  • [use this when you want to learn something new]
  1. Pure Hundreds (100, ..., 400) starts review session of words that you have already guessed at least once in a certain level (level is represented by the hundred decimal)
  • [use this when you want to practice what you've already learned in a certain level]
  1. Hundreds up to 30 (101, ..., 130) starts session of one specific lesson in a certain level (hundred decimal represents the level, the number up to 30 represents a lesson).
  • [use this when you want to learn/practice one specific lesson] """ text_vocab_interact = """ ✅ - know without thinking 🤔 - know after thinking 🧩 - know only partially ❌ - not know 🔁 - repeat the audio 📄 - display more info about the word 🔚 - end the session

If you encounter a word for the first time, it displays all the info about the word. If you encountered it at least once, info will be hidden. The frequency of certain words showing up depend on the marks you click (especially in the review sessions that also take time into account) ㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡㅡ """ text_listen = """

  1. One (1) finds and starts the lesson with the highest number in which all the words were already guessed
  2. Hundreds up to 30 (101, ..., 130) starts session of one specific lesson in a certain level """ text_listen_interact = """ ⏪ - rewind by 10 seconds ⏸️ - pause ⏩ - next track 🔁 - repeat track 🔚 - end the session

Note that listening sessions starts from 102 and reading sessions from 105.

Used resources:

  • Hongik University Language School audio files and grammar notes

  • Vocabulary audio downloaded from Naver

  • A Frequency Dictionary of Korean (book)

  • content generated by ChatGPT """

      text_links = f"""

  • Level 1 Grammar (Google Doc - grammar reference that is being used in listening/reading sessions)

  • User's stats (Google Sheet - information about user's guessings and sorted knowledge of words)

  • Vocabulary (Google Sheet - whole set of korean words)"""