lstm_sentiment_analysis

Sentiment analysis on movie reviews using LSTM neural networks

Task

Given a movie review classify it as either positive or negative, after training the agent on an adnotated dataset.

Dataset

For this task i have used the "Large Movie Review Dataset" provided by Stanford University

You can read their paper on this subject here :

   @InProceedings{maas-EtAl:2011:ACL-HLT2011,
      author    = {Maas, Andrew L.  and  Daly, Raymond E.  and  Pham, Peter T.  and  Huang, Dan  and  Ng, Andrew Y.  and  Potts, Christopher},
      title     = {Learning Word Vectors for Sentiment Analysis},
      booktitle = {Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies},
      month     = {June},
      year      = {2011},
      address   = {Portland, Oregon, USA},
      publisher = {Association for Computational Linguistics},
      pages     = {142--150},
      url       = {http://www.aclweb.org/anthology/P11-1015}
    }

Required packages

• numpy
• tensorflow
• spacy - all word embedding vectors were loaded using spacy

Usage

If you have not yet downloaded the dataset mentioned above, the script will attempt to download it for you on the first run.

First you should run train_LSTM.py. This script will preprocess the dataset and save it in a separate folder. Afterwards it will begin training the LSTM neural network. To visualize the progress in Tensorboard you can run the following command tensorboard --logdir=logs

After training the network you can run the script test_LSTM.py to see how well the network fits the test set.

The most exciting script by far is own_input_test_LSTM.py where you can enter a review as a string and see how the network classifies it.