TransPhrase: A new method for generating phrase embedding from word embedding in Chinese

作者：

Highlights：

• We propose a new composition method for phrase embedding.

• It can represent the order and weight information of component words.

• The evaluation results show it is superior to the 5 baseline composition methods.

• It is also superior to the distribution method and BERT model.

摘要

•We propose a new composition method for phrase embedding.•It can represent the order and weight information of component words.•The evaluation results show it is superior to the 5 baseline composition methods.•It is also superior to the distribution method and BERT model.

论文关键词：Phrase embedding,Word embedding,Composition method,Neural network

论文评审过程：Received 26 January 2020, Revised 24 August 2020, Accepted 25 November 2020, Available online 1 December 2020, Version of Record 10 December 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.114387