E2E Knowledge Graph Question Answering

Table of Contents

• About
• Getting Started
  • Installing the requirements
• Training
• Inferencing
• License

About

A PyTorch Implementation of Knowledge Graph Question Answering (KGQA) based on Amazon Alexa - E2E Entity Resolution and Question Answering.

The RIGEL model consists of two major components:

1. Entity Resolution: Detect Spans from Natural Language Queries and ground them to Knowledge Graph entities.

   

2. Inference Module: Learn to model relationships to traverse through the Knowlege Graph to obtain answers for the query.

   Inference Module	Example Follow Operation (2-hop)

Getting Started

Installing the requirements

Clone the repo and install requirements.txt in a Python>=3.7.0 environment, including PyTorch>=1.7.

Use the below command to recursively install all the requirements:

>> pip install -r requirements.txt

Training

The config file under ./configs/ is used to set the hyperparameters for model training and checkpoint paths.

To run train use the following command within the environment:

python .\train.py --config './configs/base.json'

Inferencing

The config file under ./configs/ is used to set the hyperparameters for innitializing model checkpoints and the question to inference.

To run train use the following command within the environment:

python .\inference.py --config './configs/base.json'

A screenshot of single query inferencing is show below:

License

This project is licensed under the MIT License. See the LICENSE file for details.