NLP notebooks

Various approaches to NLP problems using Keras and TensorFlow.

1. Sentiment analysis of IMDB reviews IMDB dataset contains labeled (25k for positive and 25k for negative) and unsupervised (50k) reviews. Models used to predict a sentiment:

• word embedding dataset GloVe. Prediction accuracy ~ 54%
• LSTM model. Prediction accuracy ~ 53% LSTM model should be fine-tuned for a better result.