Spurious Correlations in Cross-Topic Argument Mining

Terne Sasha Thorn Jakobsen, Maria Barrett, Anders Søgaard


Abstract
Recent work in cross-topic argument mining attempts to learn models that generalise across topics rather than merely relying on within-topic spurious correlations. We examine the effectiveness of this approach by analysing the output of single-task and multi-task models for cross-topic argument mining, through a combination of linear approximations of their decision boundaries, manual feature grouping, challenge examples, and ablations across the input vocabulary. Surprisingly, we show that cross-topic models still rely mostly on spurious correlations and only generalise within closely related topics, e.g., a model trained only on closed-class words and a few common open-class words outperforms a state-of-the-art cross-topic model on distant target topics.
Anthology ID:
2021.starsem-1.25
Volume:
Proceedings of *SEM 2021: The Tenth Joint Conference on Lexical and Computational Semantics
Month:
August
Year:
2021
Address:
Online
Editors:
Lun-Wei Ku, Vivi Nastase, Ivan Vulić
Venue:
*SEM
SIG:
SIGLEX
Publisher:
Association for Computational Linguistics
Note:
Pages:
263–277
Language:
URL:
https://aclanthology.org/2021.starsem-1.25
DOI:
10.18653/v1/2021.starsem-1.25
Bibkey:
Cite (ACL):
Terne Sasha Thorn Jakobsen, Maria Barrett, and Anders Søgaard. 2021. Spurious Correlations in Cross-Topic Argument Mining. In Proceedings of *SEM 2021: The Tenth Joint Conference on Lexical and Computational Semantics, pages 263–277, Online. Association for Computational Linguistics.
Cite (Informal):
Spurious Correlations in Cross-Topic Argument Mining (Thorn Jakobsen et al., *SEM 2021)
Copy Citation:
PDF:
https://preview.aclanthology.org/nschneid-patch-3/2021.starsem-1.25.pdf
Code
 terne/spurious_correlations_in_argmin