On the Systematicity of Probing Contextualized Word Representations: The Case of Hypernymy in BERT

Abhilasha Ravichander, Eduard Hovy, Kaheer Suleman, Adam Trischler, Jackie Chi Kit Cheung


Abstract
Contextualized word representations have become a driving force in NLP, motivating widespread interest in understanding their capabilities and the mechanisms by which they operate. Particularly intriguing is their ability to identify and encode conceptual abstractions. Past work has probed BERT representations for this competence, finding that BERT can correctly retrieve noun hypernyms in cloze tasks. In this work, we ask the question: do probing studies shed light on systematic knowledge in BERT representations? As a case study, we examine hypernymy knowledge encoded in BERT representations. In particular, we demonstrate through a simple consistency probe that the ability to correctly retrieve hypernyms in cloze tasks, as used in prior work, does not correspond to systematic knowledge in BERT. Our main conclusion is cautionary: even if BERT demonstrates high probing accuracy for a particular competence, it does not necessarily follow that BERT ‘understands’ a concept, and it cannot be expected to systematically generalize across applicable contexts.
Anthology ID:
2020.starsem-1.10
Volume:
Proceedings of the Ninth Joint Conference on Lexical and Computational Semantics
Month:
December
Year:
2020
Address:
Barcelona, Spain (Online)
Venues:
*SEM | COLING | SemEval
SIGs:
SIGLEX | SIGSEM
Publisher:
Association for Computational Linguistics
Note:
Pages:
88–102
Language:
URL:
https://aclanthology.org/2020.starsem-1.10
DOI:
Bibkey:
Copy Citation:
PDF:
https://aclanthology.org/2020.starsem-1.10.pdf
Code
 abhilasharavichander/probe-generalization