基于句法依赖增强图的方面级情感分析
Aspect-based sentiment analysis based on syntactic dependency enhancement graph
廖列法 1夏卫欢 2杨翌虢3
作者信息
- 1. 江西理工大学信息工程学院,江西赣州 341000;江西现代职业技术学院院长办公室,江西南昌 330095
- 2. 江西理工大学信息工程学院,江西赣州 341000
- 3. 上海大学计算机工程与科学学院,上海 200444
- 折叠
摘要
方面级情感分析旨在分析句子中特定方面的情感极性,现有研究侧重于利用图神经网络建模上下文与方面的依赖信息,忽略了对上下文中情感词及其词性的挖掘和利用.为此,提出一种基于句法依赖的增强图(syntactic dependency en-hancement graph,SDEG)模型,在原始句法依赖图上引入情感知识和词性信息,增强情感词权重和相关词性单词在上下文中的作用.使用双向长短期记忆网络和卷积神经网络捕捉句子的重点语义信息,通过图卷积神经网络建模句法依赖增强图,通过交互注意力机制生成特定方面的上下文语义和语法表示以进行情感极性分类.在多个公共基准数据集上的实验结果表明,所提模型在性能上有明显提升.
Abstract
Aspect-based sentiment analysis aims to analyze the polarity of emotion in specific aspects of sentences.The existing researches focus on using graph neural network to model contextual and aspect-dependent information,but ignore the mining and utilization of emotion words and their parts of speech in context.Therefore,the syntactic dependency enhancement graph(SDEG)model based on tactic dependency enhancement graph(SDEG)was proposed.The affective knowledge and parts of speech information were used to enhance the weight of affective words and the role of related parts of speech words in context.The bidirectional long term memory network and convolutional neural network were used to capture the key semantic information of sentences.The syntactic dependency enhancement graph was modeled through the graph convolutional neural network.The interactive attention mechanism was used to generate the contextual semantic and grammatical representation of specific aspects for emotion polarity classification.Experimental results on several common benchmark data sets show that the proposed model has significant performance improvements.
关键词
方面级情感分析/情感知识/词性/双向长短期记忆网络/卷积神经网络/图卷积神经网络/交互注意力机制Key words
aspect-based sentiment analysis/affective knowledge/part of speech/bidirectional long short-term memory net-work/convolutional neural network/graph convolution network/interactive attention mechanism引用本文复制引用
基金项目
国家自然科学基金(71462018)
国家自然科学基金(71761018)
出版年
2024
NETL
NSTL
万方数据