abdoelsayed2016 / GRG

Official PyTorch implementation of "Generate-Retrieve-Generate: A Novel Approach to Open-Domain Question Answering"

Table of Contents

• Introduction
• Requirements
• Document Generator (DG)
• Document Generator Retriever (DGR)
• Citation

Introduction

Welcome to the repository for Generator-Retriever-Generator: A Novel Approach to Open-domain Question Answering. In this work, we present the GRG approach for tackling open-domain question answering challenges.

GRG Approach

Requirements

Make sure you have the required environment set up to run the GRG project:

$ conda create -n grg
$ conda activate grg
$ pip install -r requirements.txt

Document Generator(DG)

In the Document Generator (DG) directory you'll encounter code designed to create documents using few-shot learning methodologies

Document Generator Retriever(DGR)

In the Document Generator (DG) directory, you'll uncover a retriever specifically crafted for the document generator, employing sentence transformers for enhanced performance.

Document Retriever(DR)

Will be available soon................

LLaMA Generator

Will be available soon................

Citation

If you find these codes or data useful, please consider citing our paper as:

@article{abdallah2023generator,
  title={Generator-Retriever-Generator: A Novel Approach to Open-domain Question Answering},
  author={Abdallah, Abdelrahman and Jatowt, Adam},
  journal={arXiv preprint arXiv:2307.11278},
  year={2023}
}