Skip to content

Latest commit

 

History

History
80 lines (60 loc) · 4.48 KB

README.md

File metadata and controls

80 lines (60 loc) · 4.48 KB
Raw

GProofT: A Multi-dimension Multi-round Fact Checking Framework Based on Claim Fact Extraction

This is the official code and data repository for the EMNLP 2024 Workshop FEVER AVeriTeC Shared Task paper: GProofT: A Multi-dimension Multi-round Fact Checking Framework Based on Claim Fact Extraction.

Overview

1. Requirements

we use python 3.9.19 which suits all packages

Requirements:

spaCy==3.7.5
pandas==2.2.2
openai
transformers==4.20.1
google-auth==2.31.0

2. GProofT Framework

The code for implementing GProofT framework is HERE.

Replace the following variables with your own ones in the code to execute the fact-checking process:LOAD_DATA_PATH, OUTPUT_FILE_PATH, AZURE_ENDPOINT, API_VERSION, AZURE_API_KEY, SERVICE_ACCOUNT_FILE, SEARCH_ENGINE_ID, API_KEY, where AZURE_API_KEY is the API key to use OpenAI Azure ChatGPT, and API_KEY is the API key to use Google Search API. The official documents is HERE.

3. Verdict Prediction

1. Llama_eval.py

This script is used to call large language models like LLaMA3 and LLaMA3-Instruct from Hugging Face to run zero_shot experiment and generate verdicts. For instructions on fine-tuning the model, you can refer to this repository.

2. GPT_eval.py

GPT_eval.py is used to call OpenAI model using your own API key. Replace the azure_endpoint,api_version,api_key to get response from ChatGPT. Replace input to use your own dataset.

3. Data_Processing

This directory contains code for data processing. It includes scripts to combine the verdict with retrieved evidence and restructure the data to fit a fine-tuning framework. You may use these scripts as needed if they are helpful for your specific tasks.

5. Citing this work

Please use the bibtex below for citing our paper:

@inproceedings{liu-etal-2024-gprooft,
    title = "{GP}roof{T}: A Multi-dimension Multi-round Fact Checking Framework Based on Claim Fact Extraction",
    author = "Liu, Jiayu  and
      Tang, Junhao  and
      Wang, Hanwen  and
      Xu, Baixuan  and
      Shi, Haochen  and
      Wang, Weiqi  and
      Song, Yangqiu",
    editor = "Schlichtkrull, Michael  and
      Chen, Yulong  and
      Whitehouse, Chenxi  and
      Deng, Zhenyun  and
      Akhtar, Mubashara  and
      Aly, Rami  and
      Guo, Zhijiang  and
      Christodoulopoulos, Christos  and
      Cocarascu, Oana  and
      Mittal, Arpit  and
      Thorne, James  and
      Vlachos, Andreas",
    booktitle = "Proceedings of the Seventh Fact Extraction and VERification Workshop (FEVER)",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.fever-1.14",
    pages = "118--129",
    abstract = "In the information era, the vast proliferation of online content poses significant challenges, particularly concerning the trustworthiness of these digital statements, which can have profound societal implications. Although it is possible to manually annotate and verify the authenticity of such content, the sheer volume and rapid pace of information generation render this approach impractical, both in terms of time and cost. Therefore, it is imperative to develop automated systems capable of validating online claims, ensuring that users can use the wealth of information available on the Internet effectively and reliably. Using primarily ChatGPT and the Google search API, GProofT fact checking framework generates question-answer pairs to systematically extract and verify the facts within claims. Based on the outcomes of these QA pairs, claims are subsequently labeled as Supported, Conflicted Evidence/Cherry-Picking, or Refuted. Shown by extensive experiments, GProofT Retrieval generally performs effectively in fact-checking and makes a substantial contribution to the task. Our code is released on https://github.com/HKUST-KnowComp/GProofT.",
}

6. Acknowledgement

The authors of this paper were supported by the NSFC Fund (U20B2053) from the NSFC of China, the RIF (R6020-19 and R6021-20), and the GRF (16211520 and 16205322) from RGC of Hong Kong. We also thank the support from the UGC Research Matching Grants (RMGS20EG01-D, RMGS20CR11, RMGS20CR12, RMGS20EG19, RMGS20EG21, RMGS23CR05, RMGS23EG08).