monireh2 / kg-deductive-reasoner

knowledge graph deductive reasoning using memory networks

End-To-End Memory Networks for Deductive Reasoning over Knowledge Graph

This is a modification of implementation of MemN2N model in Python for the Deductive Reasoning over Knowledge Graph as inspired by the Section 4 of the paper "End-To-End Memory Networks". It is based on Facebook's Matlab code.

[Web-based Demo](Coming soon!)

Requirements

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

$ sudo pip install -r requirements.txt

• Sample Normalized KG Dataset should be downloaded and decompressed to data:

$ mkdir data
=======

$ tar xvf sample_data_normalized.tar.xz -C data

• Sample Json Format Dataset should be downloaded and decompressed to data in case you want to run the normalization:

$ tar xvf sample_json_files.tar.xz -C data
python json_reader_normalizer.py

Usage

• To run on a knowledge graph reasoning task, use kg_reasoner_runner.py. For example,

python kg_reasoner_runner.py

The output will look like:

Using data from data/task_name/task_name
Train and test for task task_name ...
1 | train error: 0.876116 | val error: 0.75
|===================================               | 71% 0.5s

Knowledge Graph Reasoning Demo

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

python -m demo.qa_kg

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

python -m demo.qa_kg -console

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

python -m demo.qa_kg -train

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

Author

• Monireh Ebrahimi

References

• Monireh Ebrahimi, Md Kamruzzaman Sarker, Federico Bianchi, Ning Xie, Derek Doran, Pascal Hitzler "Reasoning over RDF Knowledge Bases using Deep Learning", arXiv:1811.04132[cs].

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