- Title
- What Do They Capture? - A Structural Analysis of Pre-Trained Language Models for Source Code
- Creator
- Wan, Yao; Zhao, Wei; Zhang, Hongyu; Sui, Yulei; Xu, Guandong; Jin, Hai
- Relation
- ICSE '22: 44th International Conference on Software Engineering. Proceedings of 44th International Conference on Software Engineering (Pittsburgh, PA 21-29 May, 2022) p. 2377-2388
- Publisher Link
- http://dx.doi.org/10.1145/3510003.3510050
- Publisher
- Institute of Electrical and Electronics Engineers (IEEE)
- Resource Type
- conference paper
- Date
- 2022
- Description
- Recently, many pre-trained language models for source code have been proposed to model the context of code and serve as a basis for downstream code intelligence tasks such as code completion, code search, and code summarization. These models leverage masked pre-training and Transformer and have achieved promising results. However, currently there is still little progress regarding interpretability of existing pre-trained code models. It is not clear why these models work and what feature correlations they can capture. In this paper, we conduct a thorough structural analysis aiming to provide an interpretation of pre-trained language models for source code (e.g., CodeBERT, and GraphCodeBERT) from three distinctive perspectives: (1) attention analysis, (2) probing on the word embedding, and (3) syntax tree induction. Through comprehensive analysis, this paper reveals several insightful findings that may inspire future studies: (1) Attention aligns strongly with the syntax structure of code. (2) Pre-training language models of code can preserve the syntax structure of code in the intermediate representations of each Transformer layer. (3) The pre-trained models of code have the ability of inducing syntax trees of code. Theses findings suggest that it may be helpful to incorporate the syntax structure of code into the process of pre-training for better code representations.
- Subject
- code representation; deep learning; pre-trained language model; probing, attention analysis; attention analysis; syntax tree induction
- Identifier
- http://hdl.handle.net/1959.13/1465581
- Identifier
- uon:47308
- Language
- eng
- Reviewed
- Hits: 1518
- Visitors: 1513
- Downloads: 0