Achieving Fine-grained Word Sense Disambiguation with Context Hypergraph and Sememe Hypergraph
J.Y. Liu, Haonan Zeng · 2023
Word Sense Disambiguation (WSD) is a crucial task in Natural Language Processing, aiming to identify the correct sense of a polysemous word in a given context. Current approaches often struggle to capture the intricate relationships between the target word and its possible senses. In this paper, we propose DHFM, a novel approach that leverages hypergraph representation learning to encapsulate high-order contextual dependencies and fine-grained sense distinctions. These representations are then fused to align the contextual information with the word senses accurately. Experiments on several benchmark datasets demonstrate that DHFM significantly outperforms existing state-of-the-art models. Moreover, our ablation studies validate the vital role of each component, with particular emphasis on the importance of sememe hypergraph in distinguishing nuanced senses.