Experimenting with using an LSTM language model for document embeddings; doesn't work (yet?) - View it on GitHub