融合注意力机制和Bi-LSTM的旅游评价情感分析模型  

Sentiment Analysis Model for Tourism Evaluation Integrating Attention Mechanism and Bi-LSTM

作　　者：陈玉娇 陈崇成[1] 吕贵杰 黄正睿 CHEN Yu-jiao;CHEN Chong-cheng;LYU Gui-jie;HUANG Zheng-rui(Academy of Digital China(Fujian),Fuzhou University,Fuzhou 350100,China)

机构地区：[1]福州大学数字中国研究院(福建),福建福州350100

出　　处：《软件导刊》2022年第2期43-47,共5页Software Guide

基　　金：国家重点研发计划项目(2017YFB0504202)。

摘　　要：旅游评价情感分析对了解游客偏好具有重要意义,然而现有的旅游评价情感分析模型较少突出游客情感描述的核心内容。为了更加精确地分析游客评价的情感倾向,提出一种结合词性注意力机制和双向长短期记忆神经网络(Bi-directional Long Short-Term Memory,Bi-LSTM)的旅游评价情感分析模型。该模型采用BiLSTM捕获上下文语义信息并充分利用句子中词语的词性与情感极性词之间的关系,无需依赖人工组织的情感领域词典,并使用词性注意力机制关注情感描述内容核心部分。实验结果表明,该旅游评价情感分析模型具有更高的准确率、精确度、召回率和F1得分,分别达91.84%、90.63%、92.33%、91.47%,比自注意力机制(Self Attention Mechanism,SAM)与Bi-LSTM结合的情感分析方法分别提高4.3%、1.02%、5.12%、3.08%。Sentiment analyses of online reviews for travelling are of great significance for understanding tourists’ preferences. However,the existing emotional analysis models of tourism evaluation rarely highlight the core content of tourists’ emotional description. In order to analyze emotional tendency of tourists’ evaluation more accurately,proposed a sentiment analysis model for travelling,which consists of the part-of-speech attention mechanism and Bi-directional Long Short-Term Memory(Bi-LSTM)neural network,where the model adopts Bi-LSTM to capture contextual semantic information,and makes full use of the relationship between part-of-speech of words and emotional polarity words in a sentence,and focus on the core part of the emotional description contents instead of relying on artificially organized emotional domain dic-tionaries. The results show that the sentiment analysis method proposed has higher accuracy(91.84%),precision(90.63%),recall(92.33%),and F1 score(91.47%)than that of SAM + Bi-LSTM,thus improving by4.3%,1.02%,5.12%,and 3.08%,respectively.

关 键 词：情感分析 双向长短期记忆神经网络 注意力机制 人工智能 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]