This repository is the implementation of sentiment analysis based on mLSTM model from paper Generating Reviews and Discovering Sentiment.

Python Requirements:

Python 3.5
Pytorch 0.3.0
numpy
pandas
scikit-learn
matplotlib
seaborn
unidecode
csv
json
nltk

Sentiment Visualization & Opinion Generation Tool.ipynb
This jupyter notebook provides a sentiment analysis API where you can try to visualize the sentiment expresssed in text, also generate opinions according the initial text you customize. It is based on work from @guillette and @NVIDIA.

Baseline test.ipynb
This jupyter notebook contains two baseline tests: SentiWordNet, n-gram. We test both baselines on middle scale of dataset IMDb, however they can be applied to any labelled dataset.

Data preprocessing.ipynb
This jupyter notebook contains the code for data preprocessing for several popular datasets including Twitter2016, OpinMind, CNN, Bloomberg, MPQA2.0, and IMDb. The raw dataset/corpus and the processed ones can be found in the release.