UCI Machine Learning Repository - Yelp/Amazon/IMDB reviews sentiment labelled dataset.
Text classification (Sentiment analysis) using various methods.
The purpose of this project is to assert the superiority of models who take context into account to - that is, to analyze the performance of models which consider the fact that input data might be interdependent too.
- First define baseline models : Naive Bayes and Bag of Words.
- Then, apply deep learning - convolutional neural nets (CNN's).
- Develop a Conditional Random Field (CRF) model for text classification (Sentiment Analysis).
- Toy around with LSTM encodoing - X decoding pairs (X = LSTM - Long Short Term Memory RNN / GRU - Gated Recurent Units).
- See how Skip-Thought Vectors fit into this.