摘要
电子商务突飞猛进,网购成为人们消费必不可少的渠道。网络商品评论的情感极性是获取顾客对该种类商品反馈的最直接方式,商家可以通过分析评论获取顾客对所购商品的感受,为后续销售计划变更和产品改进及时作出决策。针对CNN只能提取局部特征、RNN易导致梯度消失与爆炸的问题,提出一个结合RNN变体-GRU与MSCNN的XL-GSAtMSC模型。研究表明,在商品评论情感分类任务中,该模型各项评价指标均达到了95%,比传统模型提升了10%,既克服了传统情感领域词典的不足,又不用人为提取特征,实验证明了该模型的可行性与实用性。
With the rapid update of Internet technology and the rapid development of e-commerce,online shopping has become an indispensable way for people to buy goods.The emotional polarity of online commodity comments is the most direct way to get customers’feedback on this kind of commodities.Merchants can obtain customers’feelings on the purchased commodities by analyzing the comments,and make timely decisions for subsequent sales plan changes and product improvements.Aiming at the problem that the convolutional neural network can only extract local features and the cyclic neural network can easily lead to gradient disappearance and explosion,a XL-GSAtMSC model combining the variant of the cyclic neural network-gated cyclic unit and the multi-core jumping convolutional neural network is proposed.The research shows that in the task of classifying the emotion of commodity comments,all the evaluation indexes of this model reach 95%,which is 10% higher than the traditional model.The model not only overcomes the shortage of field dictionary in the traditional affective dictionary method,but also does not need to extract the features artificially.The results of various indicators prove the feasibility and practicability of the model.
作者
严鹏
YAN Peng(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《软件导刊》
2021年第6期75-79,共5页
Software Guide
关键词
情感分类
商品评论
深度学习
循环神经网络
卷积神经网络
emotional classification
goods comments
deep learning
recurrent neural network
convolutional neural network
