利用自注意力机制的大规模网络文档情感分析  

Large-scale web document sentiment analysis method using self attention mechanism

作　　者：夏辉丽 杨立身[2] 薛峰 XIA Hui-li;YANG Li-shen;XUE Feng(College of Computer and Artificial Intelligence,Zhengzhou University of Economics and Business,Zhengzhou 451191,China;College of Computer Science and Technology,Henan Polytechnic University,Jiaozuo 454003,China)

机构地区：[1]郑州经贸学院计算机与人工智能学院,河南郑州451191 [2]河南理工大学计算机科学与技术学院,河南焦作454003

出　　处：《计算机工程与设计》2021年第9期2642-2648,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61872126)。

摘　　要：针对社交网络文档(推文)情感分类复杂且准确度低的问题,基于MapReduce平台,提出一种利用自注意力双向分层语义模型的大规模网络文档情感分析方法。通过相似度计算对所有待分析的推文进行预归类,利用自注意力双向分层语义模型进行语义分类,准确分辨推文中词汇的情感类别,利用Hadoop框架和Hadoop分布式文件系统(HDFS)以及MapReduce编程模型实现提出的推文情感分类方法。实验结果表明,提出方法能够准确对大规模推文和词汇语义进行辨识,具有较高的计算效率,提高了情感分析的求解速度和准确度。Aiming at the complexity and low accuracy of sentiment classification of social network documents(tweets),based on the MapReduce platform,a large-scale network document sentiment analysis method using self-attention two-way hierarchical semantic model was proposed.All tweets to be analyzed were pre-categorized through similarity calculation,and the self-attention two-way hierarchical semantic model was used for semantic classification,to accurately distinguish the emotional category of the words in the tweets.The Hadoop framework,Hadoop distributed file system(HDFS)and MapReduce programming model were used to implement the proposed method of sentiment classification of tweets.Experimental results show that the proposed method can accurately identify large-scale tweets and vocabulary semantics with high computational efficiency,effectively improving the speed and accuracy of sentiment analysis.

关 键 词：MapReduce平台 情感计算 深度学习 自注意力双向分层语义模型 分布式文件系统(HDFS) 情感分类 词汇语义 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]