This repository owns a collection of assignments and notes for 'KUBIG' natural language processing session held from June to August 2021. After 3 months of session, free-themed nlp competition was held and our team took the 1st place. The lectures of the session are based on content from 'Deep learning from scratch 2'.
Weekly plan of the session and assignments are listed below.
| week | about | assignment | file |
|---|---|---|---|
| 1 | Session O.T. | - | - |
| 2 | Word Representation | pmi matrix, data crawling, wordcloud | Amazon review_n-gram |
| 3 | Word2vec | t-SNE, pre-trained models | Everytime_word2vec, t-SNE |
| 4 | Word2vec improvement | word2vec (skipgram) | News data_word2vec |
| 5 | RNN | simpleRNN | LSTM chatbot |
| 6 | Gated RNN | nlp contest prep | - |
| 7 | Text generation, Attention | nlp contest prep | - |
- Team name: MPTI (Make self-Pr Through aI)
- Project: CV generation based on GPT-3 and one-line summary with LexRank
- Result: 1st place (1/6 teams)
| date | about | to-do |
|---|---|---|
| 21.08.16 | Idea Brainstorming | Data crawling (Linkcareer, Jobkorea) |
| 21.08.19 | Data crawling/pre-processing | Data pre-processing |
| 21.08.23 | Data pre-processing | GPT-3 paper, code study |
| 21.08.26 | GPT-3 fine-tuning | Share GPT-3 fine-tuning results |
| 21.08.30 | GPT-3 fine-tuning | LexRank study, GPT-3 hyperparameter tuning |
| 21.08.31 | LexRank summarization/Final prep | Final prep |