Contextualized Word Embeddings Encode Aspects of Human-Like Word Sense Knowledge

TL;DR

This study shows that contextualized word embeddings, like BERT, reflect human-like distinctions between word senses, capturing nuances such as polysemy and homonymy through sense relatedness.

Contribution

It demonstrates that BERT embeddings encode human-like sense distinctions, bridging the gap between NLP models and human lexical understanding.

Findings

BERT sense relatedness correlates with human judgments.

Homonyms are more distant than polysemes in embedding space.

Embeddings encode nuanced sense distinctions.

Abstract

Understanding context-dependent variation in word meanings is a key aspect of human language comprehension supported by the lexicon. Lexicographic resources (e.g., WordNet) capture only some of this context-dependent variation; for example, they often do not encode how closely senses, or discretized word meanings, are related to one another. Our work investigates whether recent advances in NLP, specifically contextualized word embeddings, capture human-like distinctions between English word senses, such as polysemy and homonymy. We collect data from a behavioral, web-based experiment, in which participants provide judgments of the relatedness of multiple WordNet senses of a word in a two-dimensional spatial arrangement task. We find that participants' judgments of the relatedness between senses are correlated with distances between senses in the BERT embedding space. Homonymous senses (e.g., bat as mammal vs. bat as sports equipment) are reliably more distant from one another in the embedding space than polysemous ones (e.g., chicken as animal vs. chicken as meat). Our findings point towards the potential utility of continuous-space representations of sense meanings.