Please use this identifier to cite or link to this item:
https://scholarbank.nus.edu.sg/handle/10635/229379
DC Field | Value | |
---|---|---|
dc.title | Zero-shot Fact Verification by Claim Generation | |
dc.contributor.author | Pan, L | |
dc.contributor.author | Chen, W | |
dc.contributor.author | Xiong, W | |
dc.contributor.author | Kan, MY | |
dc.contributor.author | Wang, WY | |
dc.date.accessioned | 2022-07-28T08:48:39Z | |
dc.date.available | 2022-07-28T08:48:39Z | |
dc.date.issued | 2021-01-01 | |
dc.identifier.citation | Pan, L, Chen, W, Xiong, W, Kan, MY, Wang, WY (2021-01-01). Zero-shot Fact Verification by Claim Generation 2 : 476-483. ScholarBank@NUS Repository. | |
dc.identifier.isbn | 9781954085527 | |
dc.identifier.uri | https://scholarbank.nus.edu.sg/handle/10635/229379 | |
dc.description.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 automaticallygenerated claims that can be supported, refuted, or unverifiable from evidence from Wikipedia. QACG generates question-answer pairs from the evidence and then convert 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 zeroshot 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. | |
dc.source | Elements | |
dc.type | Conference Paper | |
dc.date.updated | 2022-07-19T07:40:33Z | |
dc.contributor.department | DEPARTMENT OF COMPUTER SCIENCE | |
dc.description.volume | 2 | |
dc.description.page | 476-483 | |
dc.published.state | Published | |
Appears in Collections: | Staff Publications Elements |
Show simple item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
2105.14682.pdf | 777.89 kB | Adobe PDF | OPEN | Published | View/Download |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.