一种基于多模型联合学习的方面情感分析方法  

ASPECT SENTIMENT ANALYSIS METHOD BASED ONMULTI-MODEL JOINT LEARNING

作　　者：杨强[1] 滕奇志[1] 卿粼波[1] 何小海[1] Yang Qiang;Teng Qizhi;Qing Linbo;He Xiaohai(Sichuan University,Chengdu 610065,Sichuan,China)

机构地区：[1]四川大学,四川成都610065

出　　处：《计算机应用与软件》2023年第8期148-154,共7页Computer Applications and Software

基　　金：家自然科学基金项目(61871278);成都市产业集群协同创新项目(2016-XT00-00015-GX);四川省科技计划项目(2018HH0143)。

摘　　要：基于方面词的情感分类是情感分析的一个重要子任务,通过学习文本上下文的信息,判别文本中特定方面的情感极性。常用的方法大多都是依赖于给定方面词和上下文词之间的语义关系来构建模型,这在一定程度上忽视了它们之间的句法关系。针对上述问题,提出一种融合注意力机制和句法信息的联合学习方法,不仅能同时学习方面词和上下文词之间的句法及语义关系,在训练过程中自动调整对方面词周围上下文词的关注度,而且同时考虑不同上下文词对方面词在句法上的修饰关系。在SemEval 2014的Restaurant数据集和Laptop数据集上进行实验,与目前最好模型相比,分类准确率和F1值都得到了较大提高。Sentiment classification based on aspect words is an important subtask of sentiment analysis.By learning the information of the context of the text,the emotion polarity of specific aspects in the text is discriminated.Most of the commonly-used methods rely on the semantic relationship between the given aspect word and the context word to build the model,which ignores the syntactic relationship between them to a certain extent.Aimed at the above problems,a joint learning method combining attention mechanism and syntactic information is proposed.This method could not only learn the syntactic and semantic relationship between aspect words and context words at the same time,automatically adjust the attention to the context words around the aspect words during the training process,but also consider the syntactic modification relations of aspect words in different context words.Experiments on SemEval 2014 s Restaurant dataset and Laptop dataset show that compared with the current optimal model,this model has greatly improved the classification accuracy and F1 value.

关 键 词：情感分析 方面词 语义关系 句法关系 联合学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]