DefSent: Sentence Embeddings using Definition Sentences

TL;DR

DefSent introduces a language-agnostic sentence embedding method leveraging dictionary definition sentences, achieving comparable or better performance on semantic similarity and evaluation tasks without relying on large NLI datasets.

Contribution

The paper presents DefSent, a novel sentence embedding approach using dictionary definitions, enabling broader language applicability without large NLI datasets.

Findings

Performs comparably on STS tasks

Slightly better on SentEval tasks

More broadly applicable across languages

Abstract

Sentence embedding methods using natural language inference (NLI) datasets have been successfully applied to various tasks. However, these methods are only available for limited languages due to relying heavily on the large NLI datasets. In this paper, we propose DefSent, a sentence embedding method that uses definition sentences from a word dictionary, which performs comparably on unsupervised semantics textual similarity (STS) tasks and slightly better on SentEval tasks than conventional methods. Since dictionaries are available for many languages, DefSent is more broadly applicable than methods using NLI datasets without constructing additional datasets. We demonstrate that DefSent performs comparably on unsupervised semantics textual similarity (STS) tasks and slightly better on SentEval tasks to the methods using large NLI datasets. Our code is publicly available at https://github.com/hpprc/defsent .