Sentiment Analysis on Yelp Dataset

This project contains the code for COMP4332 Project 1 and COMP4901K Project 2 which were on sentiment analysis on multi-label reviews (predicting stars from 1 to 5).

The data for this project is a segment of Yelp Dataset by only using 100,000 for training set and 10,000 for validation and test set respectively. The data split is illustrated in the jupyter notebook in data folder.

You can start training by running src/main.py and run inference using src/test.py which will store a prediction on the test set.

Folder Structure

data\
  Yelp_split.ipynb
results\
logs\
src\
  models\
    LSTM.py
    RCNN.py
    selfAttention.py
    LayerNorm.py
  main.py
  test.py
  load_data.py
  cls.py
config.yaml
requirements.txt

Reference

This data contains code from https://github.com/prakashpandey9/Text-Classification-Pytorch.