Natural Language Processing Notebooks

This repository is an archive of NLP notebooks, mainly in the purpose of setting baseline for each tasks.

Notebooks

01. Sentiment Analysis

(Updated 28 March, 2022)

• Sentiment binary classification on short sentences with log likelihood-ratio method.

• Task : Sequence Classification

• Dataset : Google GoEmotion

• naive_bayes_sentiment

  • Task : Binary Sentiment Classification
  • Model : Naive Bayes
  • Result : 81 % Accuracy

02. Score Clinical Patient Notes

(Updated 26 May, 2022)

• These notebooks are written for kaggle competition.

• Goal : Automatically extract feature text from human-written patient interview notes.

• Task : Token Classification (Segal et al, 2020)

• Dataset : NBME kaggle dataset

• NBME_hf

  • Uses HuggingFace API
  • Model : DistilBERT
  • Result : In Process

• NBME_pt

  • Uses PyTorch and Weights & Biases for hyperparameter tuning.
  • Model : DistilBERT
  • Result : Binary Cross Entropy 0.0139