/Sentiment_Analysis_LSTM_CNN

Sentiment analysis of IMDB Movie Review dataset with LSTM

Primary LanguagePython

#Sentiment Analysis in IMDB movie review dataset

::Sentiment analysis of IMDB Movie Review dataset with CNN::

Convolution with filter size 2,3,4.

Embedding -> Convolution -> max-pooling -> dropout -> Predictions -> Loss

Train examples = 22500 Validation examples = 2500 Embedding size = 128 Filter sizes = 3,4,5 Filters of each size = 128 Dropout keep_prob = 0.5

Epochs = 200

After 1500 steps, Train Accuracy = 100% Validation accuracy = 88%

Image courtesy:http://www.wildml.com/2015/11/understanding-convolutional-neural-networks-for-nlp/

::Sentiment analysis of IMDB Movie Review dataset with LSTM+RNN::

LSTM units = 10 Train examples = 22500 Validation examples = 2500 Embedding size = 128

After 200 steps, Training accuracy: 50%