A Dual-Pointer guided transition system for end-to-end structured sentiment analysis with global graph reasoning

作者：

Highlights：

• We design a transition-based system for end-to-end structured sentiment analysis.

• We propose a novel Dual-Pointer module into our model to better term recognition.

• We introduce a global graph reasoning mechanism for effective high-order features.

• We advance the state-of-the-art performance of structured sentiment analysis.

摘要

•We design a transition-based system for end-to-end structured sentiment analysis.•We propose a novel Dual-Pointer module into our model to better term recognition.•We introduce a global graph reasoning mechanism for effective high-order features.•We advance the state-of-the-art performance of structured sentiment analysis.

论文关键词：Natural language processing,Fine-grained sentiment analysis,Transition-based model,Structure parsing,Graph convolutional network

论文评审过程：Received 3 April 2022, Revised 28 May 2022, Accepted 5 June 2022, Available online 22 June 2022, Version of Record 22 June 2022.

论文官网地址：https://doi.org/10.1016/j.ipm.2022.102992