Skip to content

Latest commit

 

History

History
37 lines (32 loc) · 979 Bytes

README.md

File metadata and controls

37 lines (32 loc) · 979 Bytes
Raw

Relation-aware Language-Graph Transformer for Question Answering (AAAI 2023)

Official pytorch implementation of 'Relation-aware Language-Graph Transformer for Question Answering' (AAAI 2023)

Figure

Setup

  • Clone repository
git clone https://github.com/mlvlab/QAT.git
cd QAT
  • Setup conda environment
conda create -n QAT python=3.8
conda activate QAT
  • Install packages with a setup file
bash setup.sh
  • Download data

We use the question answering datasets (CommonsenseQA, OpenBookQA, and MedQA-USMLE) and their knowledge graphs. We preprocess the dataset following QA-GNN. You can download all the preprocessed data with the link.

Run Experiments

bash run_csqa.sh

Acknowledgement

This repo is built upon the QA-GNN and GSC:

https://github.com/michiyasunaga/qagnn
https://github.com/kuan-wang/graph-soft-counter