On the Evolution of Syntactic Information Encoded by BERT's Contextualized Representations
This addresses the problem of understanding knowledge evolution in fine-tuned models for NLP researchers, though it is incremental as it builds on prior studies of linguistic encoding.
The paper investigates how syntactic information encoded in BERT's representations changes during fine-tuning across six tasks, finding that it is forgotten in PoS tagging, reinforced in parsing tasks, and preserved in semantics-related tasks.
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.