Towards Debiasing Fact Verification Models
© 2019 Association for Computational Linguistics Fact verification requires validating a claim in the context of evidence. We show, however, that in the popular FEVER dataset this might not necessarily be the case. Claim-only classifiers perform competitively with top evidence-aware models. In this...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Association for Computational Linguistics
2021
|
| Online Access: | https://hdl.handle.net/1721.1/137401 |
| _version_ | 1811081226446962688 |
|---|---|
| author | Schuster, Tal Shah, Darsh Yeo, Yun Jie Serene Roberto Filizzola Ortiz, Daniel Santus, Enrico Barzilay, Regina |
| author_facet | Schuster, Tal Shah, Darsh Yeo, Yun Jie Serene Roberto Filizzola Ortiz, Daniel Santus, Enrico Barzilay, Regina |
| author_sort | Schuster, Tal |
| collection | MIT |
| description | © 2019 Association for Computational Linguistics Fact verification requires validating a claim in the context of evidence. We show, however, that in the popular FEVER dataset this might not necessarily be the case. Claim-only classifiers perform competitively with top evidence-aware models. In this paper, we investigate the cause of this phenomenon, identifying strong cues for predicting labels solely based on the claim, without considering any evidence. We create an evaluation set that avoids those idiosyncrasies. The performance of FEVER-trained models significantly drops when evaluated on this test set. Therefore, we introduce a regularization method which alleviates the effect of bias in the training data, obtaining improvements on the newly created test set. This work is a step towards a more sound evaluation of reasoning capabilities in fact verification models. |
| first_indexed | 2024-09-23T11:43:27Z |
| format | Article |
| id | mit-1721.1/137401 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T11:43:27Z |
| publishDate | 2021 |
| publisher | Association for Computational Linguistics |
| record_format | dspace |
| spelling | mit-1721.1/1374012021-11-05T03:24:30Z Towards Debiasing Fact Verification Models Schuster, Tal Shah, Darsh Yeo, Yun Jie Serene Roberto Filizzola Ortiz, Daniel Santus, Enrico Barzilay, Regina © 2019 Association for Computational Linguistics Fact verification requires validating a claim in the context of evidence. We show, however, that in the popular FEVER dataset this might not necessarily be the case. Claim-only classifiers perform competitively with top evidence-aware models. In this paper, we investigate the cause of this phenomenon, identifying strong cues for predicting labels solely based on the claim, without considering any evidence. We create an evaluation set that avoids those idiosyncrasies. The performance of FEVER-trained models significantly drops when evaluated on this test set. Therefore, we introduce a regularization method which alleviates the effect of bias in the training data, obtaining improvements on the newly created test set. This work is a step towards a more sound evaluation of reasoning capabilities in fact verification models. 2021-11-04T19:16:09Z 2021-11-04T19:16:09Z 2019 2020-12-01T16:49:55Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/137401 Schuster, Tal, Shah, Darsh, Yeo, Yun Jie Serene, Roberto Filizzola Ortiz, Daniel, Santus, Enrico et al. 2019. "Towards Debiasing Fact Verification Models." EMNLP-IJCNLP 2019 - 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing, Proceedings of the Conference. en 10.18653/V1/D19-1341 EMNLP-IJCNLP 2019 - 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing, Proceedings of the Conference Creative Commons Attribution 4.0 International license https://creativecommons.org/licenses/by/4.0/ application/pdf Association for Computational Linguistics Association for Computational Linguistics |
| spellingShingle | Schuster, Tal Shah, Darsh Yeo, Yun Jie Serene Roberto Filizzola Ortiz, Daniel Santus, Enrico Barzilay, Regina Towards Debiasing Fact Verification Models |
| title | Towards Debiasing Fact Verification Models |
| title_full | Towards Debiasing Fact Verification Models |
| title_fullStr | Towards Debiasing Fact Verification Models |
| title_full_unstemmed | Towards Debiasing Fact Verification Models |
| title_short | Towards Debiasing Fact Verification Models |
| title_sort | towards debiasing fact verification models |
| url | https://hdl.handle.net/1721.1/137401 |
| work_keys_str_mv | AT schustertal towardsdebiasingfactverificationmodels AT shahdarsh towardsdebiasingfactverificationmodels AT yeoyunjieserene towardsdebiasingfactverificationmodels AT robertofilizzolaortizdaniel towardsdebiasingfactverificationmodels AT santusenrico towardsdebiasingfactverificationmodels AT barzilayregina towardsdebiasingfactverificationmodels |