We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: What do you mean, BERT? Assessing BERT as a Distributional Semantics Model

Abstract: Contextualized word embeddings, i.e. vector representations for words in context, are naturally seen as an extension of previous noncontextual distributional semantic models. In this work, we focus on BERT, a deep neural network that produces contextualized embeddings and has set the state-of-the-art in several semantic tasks, and study the semantic coherence of its embedding space. While showing a tendency towards coherence, BERT does not fully live up to the natural expectations for a semantic vector space. In particular, we find that the position of the sentence in which a word occurs, while having no meaning correlates, leaves a noticeable trace on the word embeddings and disturbs similarity relationships.
Subjects: Computation and Language (cs.CL)
Journal reference: Proceedings of the Society for Computation in Linguistics: Vol. 3 (2020), Article 34
DOI: 10.7275/t778-ja71
Cite as: arXiv:1911.05758 [cs.CL]
  (or arXiv:1911.05758v2 [cs.CL] for this version)

Submission history

From: Timothee Mickus [view email]
[v1] Wed, 13 Nov 2019 19:04:42 GMT (102kb,D)
[v2] Fri, 8 May 2020 14:40:36 GMT (102kb,D)

Link back to: arXiv, form interface, contact.