Improving interpretability of word embeddings by generating definition and usage
作者：
Highlights：
• A model with gated mechanism is proposed for generating context-aware definitions.
• Scaled dot-product attention captures the interaction between contexts and words.
• ELMo embeddings are used to compensate for the drawbacks of word embeddings.
• Usage modeling is proposed to further improve the interpretability of embeddings.
• Our definition model with multi-task learning achieves significant improvement.
摘要
•A model with gated mechanism is proposed for generating context-aware definitions.•Scaled dot-product attention captures the interaction between contexts and words.•ELMo embeddings are used to compensate for the drawbacks of word embeddings.•Usage modeling is proposed to further improve the interpretability of embeddings.•Our definition model with multi-task learning achieves significant improvement.
论文关键词：Word embeddings interpretability,Definition modeling,Definition generation,Usage modeling,Usage generation
论文评审过程：Received 30 August 2019, Revised 30 April 2020, Accepted 4 June 2020, Available online 16 June 2020, Version of Record 1 July 2020.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.113633
