基于层次注意力机制和门机制的属性级别情感分析(Aspect-level Sentiment Analysis Based on Hierarchical Attention and Gate Networks)

Chao Feng (冯超), Haihui Li (黎海辉), Hongya Zhao (赵洪雅), Yun Xue (薛云), Jingyao Tang (唐靖尧)


Abstract
近年来,作为细粒度的属性级别情感分析在商业界和学术界受到越来越多的关注,其目的在于识别一个句子中多个属性词所对应的情感极性。目前,在解决属性级别情感分析问题的绝大多数工作都集中在注意力机制的设计上,以此突出上下文和属性词中不同词对于属性级别情感分析的贡献,同时使上下文和属性词之间相互关联。本文提出使用层次注意力机制和门机制处理属性级别情感分析任务,在得到属性词的隐藏状态之后,通过注意力机制得到属性词新的表示,然后利用属性词新的表示和注意力机制进一步得到上下文新的表示,层次注意力机制的设计使得上下文和属性词的表达更加准确;同时通过门机制选择对属性词而言上下文中有用的信息,以此丰富上下文的表达,在SemEval 2014 Task4和Twitter数据集上的实验结果表明本文提出模型的有效性。
Anthology ID:
2020.ccl-1.64
Volume:
Proceedings of the 19th Chinese National Conference on Computational Linguistics
Month:
October
Year:
2020
Address:
Haikou, China
Editors:
Maosong Sun (孙茂松), Sujian Li (李素建), Yue Zhang (张岳), Yang Liu (刘洋)
Venue:
CCL
SIG:
Publisher:
Chinese Information Processing Society of China
Note:
Pages:
688–697
Language:
Chinese
URL:
https://aclanthology.org/2020.ccl-1.64
DOI:
Bibkey:
Cite (ACL):
Chao Feng, Haihui Li, Hongya Zhao, Yun Xue, and Jingyao Tang. 2020. 基于层次注意力机制和门机制的属性级别情感分析(Aspect-level Sentiment Analysis Based on Hierarchical Attention and Gate Networks). In Proceedings of the 19th Chinese National Conference on Computational Linguistics, pages 688–697, Haikou, China. Chinese Information Processing Society of China.
Cite (Informal):
基于层次注意力机制和门机制的属性级别情感分析(Aspect-level Sentiment Analysis Based on Hierarchical Attention and Gate Networks) (Feng et al., CCL 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.ccl-1.64.pdf