Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning

Forrest Davis, Marten van Schijndel


Abstract
A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processes within a language, rather than just non-linguistic model biases, could obscure underlying linguistic knowledge. We tested this claim by exploring a single phenomenon in four languages: English, Chinese, Spanish, and Italian. While human behavior has been found to be similar across languages, we find cross-linguistic variation in model behavior. We show that competing processes in a language act as constraints on model behavior and demonstrate that targeted fine-tuning can re-weight the learned constraints, uncovering otherwise dormant linguistic knowledge in models. Our results suggest that models need to learn both the linguistic constraints in a language and their relative ranking, with mismatches in either producing non-human-like behavior.
Anthology ID:
2021.acl-long.93
Volume:
Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)
Month:
August
Year:
2021
Address:
Online
Editors:
Chengqing Zong, Fei Xia, Wenjie Li, Roberto Navigli
Venues:
ACL | IJCNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1159–1171
Language:
URL:
https://aclanthology.org/2021.acl-long.93
DOI:
10.18653/v1/2021.acl-long.93
Bibkey:
Cite (ACL):
Forrest Davis and Marten van Schijndel. 2021. Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1159–1171, Online. Association for Computational Linguistics.
Cite (Informal):
Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning (Davis & van Schijndel, ACL-IJCNLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.acl-long.93.pdf
Video:
 https://aclanthology.org/2021.acl-long.93.mp4
Code
 forrestdavis/ImplicitCausality