Integrating Distributional Lexical Contrast into Word Embeddings for Antonym-Synonym Distinction

TL;DR

This paper introduces a new vector representation that incorporates lexical contrast to improve word embeddings, significantly enhancing the ability to distinguish antonyms from synonyms and outperforming existing models.

Contribution

The paper presents a novel method for integrating lexical contrast into distributional vectors and embedding training, improving semantic distinction capabilities.

Findings

Outperforms standard models in antonym-synonym distinction with 0.66-0.76 precision

Enhances word similarity prediction on the SimLex-999 dataset

Outperforms state-of-the-art models in lexical contrast tasks

Abstract

We propose a novel vector representation that integrates lexical contrast into distributional vectors and strengthens the most salient features for determining degrees of word similarity. The improved vectors significantly outperform standard models and distinguish antonyms from synonyms with an average precision of 0.66-0.76 across word classes (adjectives, nouns, verbs). Moreover, we integrate the lexical contrast vectors into the objective function of a skip-gram model. The novel embedding outperforms state-of-the-art models on predicting word similarities in SimLex-999, and on distinguishing antonyms from synonyms.