README.md

CFEVER-data

Introduction to CFEVER

This repository contains the dataset for our AAAI 2024 paper, "CFEVER: A Chinese Fact Extraction and VERification Dataset". (Paper link will be provided soon.)

Leaderboard website

Please visit https://ikmlab.github.io/CFEVER to check the leaderboard of CFEVER.

Repository structure

CFEVER-data
├── data
│   ├── dev.jsonl # CFEVER development set
│   ├── test.jsonl # CFEVER test set without labels and evidence
│   └── train.jsonl # CFEVER training set
├── LICENSE
├── README.md
└── sample_submission.jsonl # sample submission file of the test set

Getting started

• Download this repository

git clone https://github.com/IKMLab/CFEVER-data.git
cd CFEVER-data

• Download the processed Wikipedia fact database from this google drive link and unzip it.

unzip wiki-pages.zip

• Then you will get a folder named wiki-pages containing 24 jsonl files. Each file contains the 50,000 processed Wikipedia pages.
  • In each jsonl file, each line is a json object representing a Wikipedia page. The json object has the following fields:
    • id: the Wikipedia page name
    • text: the processed text of the Wikipedia article
    • lines: the processed text of the Wikipedia article including the sentence numbers

Evaluation

• Please refer to our codebase: https://github.com/IKMLab/CFEVER-baselines/?tab=readme-ov-file#evaluations

Submission

• Please include three fields (necessary) in the prediction file for each claim in the test set.
  • id
  • predicted_label
  • predicted_evidence
• The id field has been already included in the test set. Please do not change the order.
• The predicted_label should be one of supports, refutes, or NOT ENOUGH INFO.
• The predicted_evidence should be a list of evidence sentences, where each evidence sentence is represented by a list of [page_id, line_number]. For example:

# One evidence sentence for the claim
{
    "id": 1,
    "predicted_label": "REFUTES",
    "predicted_evidence": [
        ["page_id_2", 2],
    ]
}

# Two evidence sentences for the claim
{
    "id": 1,
    "predicted_label": "SUPPORTS",
    "predicted_evidence": [
        ["page_id_1", 1],
        ["page_id_2", 2],
    ]
}

# The claim cannot be verified
{
    "id": 1,
    "predicted_label": "NOT ENOUGH INFO",
    "predicted_evidence": None
}

• After creating the prediction file, please email the file to [email protected] with a brief description of your method. We will evaluate your submission and update the leaderboard.
• A randomly generated submission file can be found here.
  • Note that claim is not necessary to be included in the submission file.
• You can also check the prediction example for the development set and follow the evaluation steps from our CFEVER-baselines repo.

Reference

If you find our work useful, please cite our paper.

@article{Lin_Lin_Yeh_Li_Hu_Hsu_Lee_Kao_2024,
    title = {CFEVER: A Chinese Fact Extraction and VERification Dataset},
    author = {Lin, Ying-Jia and Lin, Chun-Yi and Yeh, Chia-Jen and Li, Yi-Ting and Hu, Yun-Yu and Hsu, Chih-Hao and Lee, Mei-Feng and Kao, Hung-Yu},
    doi = {10.1609/aaai.v38i17.29825},
    journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
    month = {Mar.},
    number = {17},
    pages = {18626-18634},
    url = {https://ojs.aaai.org/index.php/AAAI/article/view/29825},
    volume = {38},
    year = {2024},
    bdsk-url-1 = {https://ojs.aaai.org/index.php/AAAI/article/view/29825},
    bdsk-url-2 = {https://doi.org/10.1609/aaai.v38i17.29825}
}