摘要
方面级情感分类旨在判断句子中每个具体方面的情感极性.传统的注意力机制模型可能会给句子中重要情感词分配过低的注意力权重,而且很少考虑上下文与方面词的交互信息.针对第1个问题,本文改进了传统的输入方式,以方面词为界限,将句子划分成包含方面词的上文、方面词和包含方面词的下文3部分作为输入,分别提取上文或下文中的重要情感特征.针对第2个问题,本文提出了词级交互注意力机制,分别学习上文与方面词、下文与方面词的词级交互,得到特定于方面的上文表示和下文表示向量,最后将它们拼接得到特定于方面的上下文表示向量,作为方面级情感分类特征.通过在3个标准数据集上的实验证明,本文的模型性能优于基线模型.
The purpose of aspect level sentiment classification is to determine the sentiment polarity of each specific aspect in a sentence.The traditional attention mechanism model may assign too low attention weight to important sentimental words in sentences,and seldom consider the interaction information between context and aspect words.Aiming at the first problem,this paper improves the traditional input method.This paper divides the sentence into three parts,namely,the above of the sentence containing the aspect word,the aspect word and the following containing the aspect word.We extract the important emotional features from the above or following.To solve the second problem,this paper proposes a word-level interactive attention mechanism.By learning the word level interaction between the above and the aspect words,the following and the aspect words,we can get the aspect specific expression vectors of the above and the following,and finally splice them to obtain the aspect specific context representation vector as the aspect level sentiment classification feature.Experiments on three standard datasets show that the performance of the proposed model is better than that of the baseline model.
作者
杨春霞
瞿涛
李欣栩
YANG Chun-xia;QU Tao;LI Xin-xu(Nanjing University of Information Science&Technology Automation Institute,Nanjing 210044,China;Jiangsu Key Laboratory of Big Data Analysis Technology,Nanjing 210044,China;Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology,Nanjing 210044,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2022年第7期1432-1437,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61273229)资助.
关键词
方面级情感分类
上下文
方面词
双向长短期记忆网络
词级交互注意力机制
aspect level sentiment classification
context
aspect words
bi-directional long short term memory
world-level interactive attention mechanism
