Aspect-based sentiment analysis with graph convolution over syntactic dependencies

作者：

Highlights：

• Graph convolutional network outperforms recurrent networks and long short-term memory on aspect-based sentiment analysis.

• Graph convolution over dependency parse trees outperforms the same approach over a flat sequence representation of sentences.

• Graph convolutional network effectively utilises context to determine the intended sentiment of inherently negative concepts.

摘要

•Graph convolutional network outperforms recurrent networks and long short-term memory on aspect-based sentiment analysis.•Graph convolution over dependency parse trees outperforms the same approach over a flat sequence representation of sentences.•Graph convolutional network effectively utilises context to determine the intended sentiment of inherently negative concepts.

论文关键词：Sentiment analysis,Natural language processing,Dependency parsing,Neural network,Graph convolutional network

论文评审过程：Received 16 December 2020, Revised 5 June 2021, Accepted 3 August 2021, Available online 9 August 2021, Version of Record 11 August 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102138