Homonym Identification using BERT -- Using a Clustering Approach

TL;DR

This paper explores the use of BERT embeddings combined with clustering algorithms to identify homonyms in text, aiming to improve coarse-grained word sense disambiguation by leveraging contextual information.

Contribution

It introduces a clustering approach using BERT embeddings for homonym identification, demonstrating the potential of contextual embeddings over traditional methods.

Findings

BERT embeddings effectively capture context for homonym detection

Clustering algorithms can distinguish different senses in embedding space

Visualization supports the feasibility of the clustering approach

Abstract

Homonym identification is important for WSD that require coarse-grained partitions of senses. The goal of this project is to determine whether contextual information is sufficient for identifying a homonymous word. To capture the context, BERT embeddings are used as opposed to Word2Vec, which conflates senses into one vector. SemCor is leveraged to retrieve the embeddings. Various clustering algorithms are applied to the embeddings. Finally, the embeddings are visualized in a lower-dimensional space to understand the feasibility of the clustering process.