On the evolution of syntactic information encoded by BERT’s contextualized representations

Laura Pérez-Mayos, Roberto Carlini, Miguel Ballesteros, Leo Wanner


Abstract
The adaptation of pretrained language models to solve supervised tasks has become a baseline in NLP, and many recent works have focused on studying how linguistic information is encoded in the pretrained sentence representations. Among other information, it has been shown that entire syntax trees are implicitly embedded in the geometry of such models. As these models are often fine-tuned, it becomes increasingly important to understand how the encoded knowledge evolves along the fine-tuning. In this paper, we analyze the evolution of the embedded syntax trees along the fine-tuning process of BERT for six different tasks, covering all levels of the linguistic structure. Experimental results show that the encoded syntactic information is forgotten (PoS tagging), reinforced (dependency and constituency parsing) or preserved (semantics-related tasks) in different ways along the fine-tuning process depending on the task.
Anthology ID:
2021.eacl-main.191
Volume:
Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume
Month:
April
Year:
2021
Address:
Online
Venue:
EACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
2243–2258
Language:
URL:
https://aclanthology.org/2021.eacl-main.191
DOI:
10.18653/v1/2021.eacl-main.191
Bibkey:
Cite (ACL):
Laura Pérez-Mayos, Roberto Carlini, Miguel Ballesteros, and Leo Wanner. 2021. On the evolution of syntactic information encoded by BERT’s contextualized representations. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 2243–2258, Online. Association for Computational Linguistics.
Cite (Informal):
On the evolution of syntactic information encoded by BERT’s contextualized representations (Pérez-Mayos et al., EACL 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.eacl-main.191.pdf
Data
GLUEPenn Treebank