Determine how positive a Movie Review is by analyising the text.
This project combines 1 Dimensional Convolutions, Word Embeddings learned in an unsupervised manner and Long Short Term Memory networks to predict if a movie review is positive.
I used the pretrained GloVe (Global Vectors for Word Representation) word embeddings by Stanford to greatly reduce training time.
A set of 25,000 highly polar movie reviews for training, and 25,000 for testing.
| Input |
|---|
| Word Embedding |
| Convolution 1D |
| MaxPooling |
| Convolution 1D |
| MaxPooling |
| Convolution 1D |
| LSTM |
| Global MaxPooling |
| Fully Connected |
Download the movie review dataset from here and extract to data directory.
Download GloVe word embedding model from here (822MB)
Place the glove6B.100d.txt file in models directory.