End-To-End Memory Networks in Tensorflow applied to Sequential Modelling

This is a Tensorflow implementation of the End-to-End Memory Network applied to Sequential Modelling of Facebook comments.

Data

The data is already available in data.tar.gz and must be uncompressed in the data/ folder.

To train the Memory Network, simply run the following command:

$ python3 main.py

Different options can be set as arguments and are described as comments of the main.py file.

We achieve the following performance:

Network	Perplexity
MN	121.45

Rémi Vincent / @remeus