基于SVM和LSTM两种模型的商品评论情感分析研究
|
摘要
随着网购的盛行,商品评论数量急剧增长,内容也越来越五花八门。如何高效挖掘处理这些评论是一件非常有价值的事情。对商品评论做情感分析是关于这些评论研究的一个重要方向。现阶段在情感分析研究中最常用的有基于机器学习的方法和基于情感知识分析的方法。本文主要采用机器学习中的SVM方法和深度学习中的LSTM方法分别对从京东网站爬取的商品评论进行模型搭建,然后对比分析。由于LSTM能够保持长期的记忆性,它很好地克服在SVM分类中每个句子的词向量求平均丢失了句子词语之间的顺序信息的缺点,保留了词与词之间的语义信息(如词序信息、上下文信息等),并且通过复杂的非线性计算更好地提取词向量中隐藏的情感信息。因此使用LSTM方法准确率比SVM方法提高不少,在情感分析上表现出非常好的效果。
With the popularity of online shopping, the number of product reviews has increased dramatically, and its contents are becoming more and more diverse. How to efficiently mine these reviews is a very valuable thing.Emotional analysis of product reviews is an important aspect of these reviews. The most commonly used methods in sentiment analysis are machine-based learning and sentiment knowledge analysis at present. In this paper, SVM method in machine learning and LSTM method in depth learning are used to model the product reviews crawled from Jingdong website. Because LSTM method can maintain long-term memory, it can overcome the shortcoming of losing the order information between words in each sentence by SVM method, so the accuracy of LSTM method in test set is much higher than that of SVM method.
With the popularity of online shopping, the number of product reviews has increased dramatically, and its contents are becoming more and more diverse. How to efficiently mine these reviews is a very valuable thing.Emotional analysis of product reviews is an important aspect of these reviews. The most commonly used methods in sentiment analysis are machine-based learning and sentiment knowledge analysis at present. In this paper, SVM method in machine learning and LSTM method in depth learning are used to model the product reviews crawled from Jingdong website. Because LSTM method can maintain long-term memory, it can overcome the shortcoming of losing the order information between words in each sentence by SVM method, so the accuracy of LSTM method in test set is much higher than that of SVM method.
引文
[1]钟将,杨思源,孙启干.基于文本分类的商品评价情感分析[J].计算机应用,2014,34(8):2317-2321.
[2]张紫琼,叶强,李一军.互联网商品评论情感分析研究综述[J].管理科学学报,2010,13(6):84-96.
[3]Scholz T,Conrad S.Linguistic Sentiment Features for Newspaper Opinion Mining[M]//Natural Language Processing and Information Systems.Springer Berlin Heidelberg,2013:272-277.
[4]赵妍妍,秦兵,刘挺.文本情感分析[J].软件学报,2010,21(8):1834-1848.
[5]樊康新.基于SVM的网络文本情感分类系统的研究与设计[J].计算机时代,2015(12):34-37.
[6]Jun,Liang,et al."Polarity Shifting and LSTM Based Recursive Networks for Sentiment Analysis."Journal of Chinese Information Processing(2015).
[7]Harrington P.Machine Learning in Action[M].Manning Publications Co,2012.
[8]樊康新.基于SVM的网络文本情感分类系统的研究与设计[J].计算机时代,2015(12):34-37.
[9]Mikolov T,Chen K,Corrado G,et al.Efficient Estimation of Word Representations in Vector Space[J].Computer Science,2013.
[10]陈葛恒.基于极性转移和双向LSTM的文本情感分析[J].信息技术,2018(2):149-152.
[11]Abadi M.TensorFlow:learning functions at scale[J].Acm Sigplan Notices,2016,51(9):1-1.
[12]Zhang D,Xu H,Su Z,et al.Chinese comments sentiment classification based on word2vec and SVM perf[J].Expert Systems with Applications,2015,42(4):1857-1863.
[13]Tang D,Qin B,Feng X,et al.Effective LSTMs for TargetDependent Sentiment Classification[J].Computer Science,2015.
[2]张紫琼,叶强,李一军.互联网商品评论情感分析研究综述[J].管理科学学报,2010,13(6):84-96.
[3]Scholz T,Conrad S.Linguistic Sentiment Features for Newspaper Opinion Mining[M]//Natural Language Processing and Information Systems.Springer Berlin Heidelberg,2013:272-277.
[4]赵妍妍,秦兵,刘挺.文本情感分析[J].软件学报,2010,21(8):1834-1848.
[5]樊康新.基于SVM的网络文本情感分类系统的研究与设计[J].计算机时代,2015(12):34-37.
[6]Jun,Liang,et al."Polarity Shifting and LSTM Based Recursive Networks for Sentiment Analysis."Journal of Chinese Information Processing(2015).
[7]Harrington P.Machine Learning in Action[M].Manning Publications Co,2012.
[8]樊康新.基于SVM的网络文本情感分类系统的研究与设计[J].计算机时代,2015(12):34-37.
[9]Mikolov T,Chen K,Corrado G,et al.Efficient Estimation of Word Representations in Vector Space[J].Computer Science,2013.
[10]陈葛恒.基于极性转移和双向LSTM的文本情感分析[J].信息技术,2018(2):149-152.
[11]Abadi M.TensorFlow:learning functions at scale[J].Acm Sigplan Notices,2016,51(9):1-1.
[12]Zhang D,Xu H,Su Z,et al.Chinese comments sentiment classification based on word2vec and SVM perf[J].Expert Systems with Applications,2015,42(4):1857-1863.
[13]Tang D,Qin B,Feng X,et al.Effective LSTMs for TargetDependent Sentiment Classification[J].Computer Science,2015.