English

Zero-shot Fact Verification by Claim Generation

Computation and Language 2021-06-01 v1 Artificial Intelligence

Abstract

Neural models for automated fact verification have achieved promising results thanks to the availability of large, human-annotated datasets. However, for each new domain that requires fact verification, creating a dataset by manually writing claims and linking them to their supporting evidence is expensive. We develop QACG, a framework for training a robust fact verification model by using automatically generated claims that can be supported, refuted, or unverifiable from evidence from Wikipedia. QACG generates question-answer pairs from the evidence and then converts them into different types of claims. Experiments on the FEVER dataset show that our QACG framework significantly reduces the demand for human-annotated training data. In a zero-shot scenario, QACG improves a RoBERTa model's F1 from 50% to 77%, equivalent in performance to 2K+ manually-curated examples. Our QACG code is publicly available.

Keywords

Cite

@article{arxiv.2105.14682,
  title  = {Zero-shot Fact Verification by Claim Generation},
  author = {Liangming Pan and Wenhu Chen and Wenhan Xiong and Min-Yen Kan and William Yang Wang},
  journal= {arXiv preprint arXiv:2105.14682},
  year   = {2021}
}

Comments

ACL-IJCNLP 2021 (main conference, short paper)

R2 v1 2026-06-24T02:38:34.174Z