What Does BERT Look at? An Analysis of BERT's Attention

Kevin Clark, Urvashi Khandelwal, Omer Levy, Christopher D. Manning

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Large pre-trained neural networks such as BERT have had great recent success in NLP, motivating a growing body of research investigating what aspects of language they are able to learn from unlabeled data. Most recent analysis has focused on model outputs (e.g., language model surprisal) or internal vector representations (e.g., probing classifiers). Complementary to these works, we propose methods for analyzing the attention mechanisms of pre-trained models and apply them to BERT. BERT's attention heads exhibit patterns such as attending to delimiter tokens, specific positional offsets, or broadly attending over the whole sentence, with heads in the same layer often exhibiting similar behaviors. We further show that certain attention heads correspond well to linguistic notions of syntax and coreference. For example, we find heads that attend to the direct objects of verbs, determiners of nouns, objects of prepositions, and coreferent mentions with remarkably high accuracy. Lastly, we propose an attention-based probing classifier and use it to further demonstrate that substantial syntactic information is captured in BERT's attention.
Original languageEnglish
Title of host publicationProceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP
EditorsTal Linzen, Grzegorz Chrupała, Yonatan Belinkov, Dieuwke Hupkes
Place of PublicationFlorence, Italy
PublisherAssociation for Computational Linguistics
Pages276-286
Number of pages11
ISBN (Electronic)978-1-950737-30-7
DOIs
StatePublished - 1 Aug 2019
Externally publishedYes
EventThe 2nd ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP - Florence, Italy
Duration: 1 Aug 20191 Aug 2019
Conference number: 2

Workshop

WorkshopThe 2nd ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP
Country/TerritoryItaly
CityFlorence
Period1/08/191/08/19

Fingerprint

Dive into the research topics of 'What Does BERT Look at? An Analysis of BERT's Attention'. Together they form a unique fingerprint.

Cite this