Text Generator (LSTM + Keras)

The code builds an LSTM to generate text using Keras. Watch this video by Ajay Halthor to see a live demo and understand the intuition behind Recurrent Neural Networks.

Dependencies

• python=3.6.0
• numpy=1.14.0
• nltk=3.2.3
• Tensorflow=1.5.0
• Keras=2.1.3

Files

• notebook: All the code for generating the LSTM
• hdf5 file: saved model. You can use it to continue training for better results.

Execution

Open this project directory in terminal and type

$ jupyter notebook

Recommend install jupyter through anaconda here.

References

• [1] Deep Learning book Ch.10 on Sequence modelling: https://www.deeplearningbook.org/contents/rnn.html
• [2] Colah’s blog on LSTMs: http://colah.github.io/posts/2015-08-Understanding-LSTMs/
• [3] Andrej Karpathy’s Blog + Code (You can probably understand more from this now!): http://karpathy.github.io/2015/05/21/rnn-effectiveness/