基于神经网络的多注意属性情感分析  

Multi-attention aspect-level sentiment analysis based on neural network

作　　者：陈闰雪 彭龑 徐莲 CHEN Run-xue;PENG Yan;XU Lian(School of Computer,Sichuan University of Science and Engineering,Zigong 643000,China)

机构地区：[1]四川轻化工大学计算机学院,四川自贡643000

出　　处：《计算机工程与设计》2020年第10期2963-2968,共6页Computer Engineering and Design

基　　金：四川省科技厅科技支撑计划基金项目(19ZDYF1078);自贡市重点科技计划基金项目(2018GYCX33)。

摘　　要：针对传统神经网络无法准确捕捉多个目标属性情感特征的问题,提出一种基于神经网络的多注意属性情感分析模型。用双向长短时记忆提取上下文和目标属性之间关联,用位置编码机制捕捉与目标属性相邻词的重要信息,结合位置编码和强注意力机制挖掘上下文和目标属性进行情感分析,在SemEval 2014和twitter数据集进行实验。实验结果表明,该模型在laptop、restaurant和twitter数据集上的准确率分别是72.45%、80.17%和73.39%,相比传统神经网络准确率得到了提高。Aiming at the problem that traditional neural network can not accurately capture the emotional characteristics of multiple target attributes,a multi-attention aspect-level sentiment analysis model based on neural network was proposed.Bidirectional long short-term memory was used to extract the association between the context and the target attribute.The position-coding mechanism was used to capture important information about words adjacent to target attributes.Position-coding and strong attention mechanisms were combined to mine context and target attributes for sentiment analysis.Experiment were carried out on the SemEval 2014 and twitter data sets.Experimental results show that the accuracies of the proposed model on the laptop,restaurant and twitter data sets are 72.45%,80.17%and 73.39%,respectively,which are better than that of the traditional neural network.

关 键 词：多注意 属性情感分析 神经网络 双向长短时记忆 位置编码 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]