End-To-End Memory Networks for Question Answering

This is an implementation of MemN2N model in Python for the bAbI question-answering tasks as shown in the Section 4 of the paper "End-To-End Memory Networks". It is based on Facebook's Matlab code.

Requirements

• Python 2.7
• Numpy, Flask (only for web-based demo) can be installed via pip:

$ sudo pip install -r requirements.txt

• bAbI dataset should be downloaded to data/tasks_1-20_v1-2:

$ wget -qO- http://www.thespermwhale.com/jaseweston/babi/tasks_1-20_v1-2.tar.gz | tar xvz -C data

Usage

• To run on a single task, use babi_runner.py with -t followed by task's id. For example,

python babi_runner.py -t 1

The output will look like:

Using data from data/tasks_1-20_v1-2/en
Train and test for task 1 ...
1 | train error: 0.876116 | val error: 0.75
|===================================               | 71% 0.5s

• To run on 20 tasks:

python babi_runner.py -a

• To train using all training data from 20 tasks, use the joint mode:

python babi_runner.py -j

Question Answering Demo

• In order to run the Web-based demo using the pretrained model memn2n_model.pklz in trained_model/, run:

python -m demo.qa

• Alternatively, you can try the console-based demo:

python -m demo.qa -console

• The pretrained model memn2n_model.pklz can be created by running:

python -m demo.qa -train

• To show all options, run python -m demo.qa -h

Benchmarks

See the results here.

Author

Vinh Khuc

Future Plans

• Port to TensorFlow/Keras
• Support Python 3

References

• Sainbayar Sukhbaatar, Arthur Szlam, Jason Weston, Rob Fergus, "End-To-End Memory Networks", arXiv:1503.08895 [cs.NE].