基于双向LSTM的动态情感词典构建方法研究
|
摘要
文本情感分析最基础且最关键的一个环节就是构建一个高质量情感词典.为克服传统的情感词典中词汇所表达出的情感倾向不变性等问题的不足,本文基于三层神经网络结构构建动态情感词典.第一层通过改进的CBOW神经网络提取含有情感信息的特征;第二层通过双向LSTM神经网络,利用二叉语义依存结构模型提取出二叉语义依存路径特征;第三层在前两层获得的情感特征和语义特征基础上,将中心词信息和词汇到中心词的距离两个特征一起组成当前词的特征,然后,对双向LSTM神经网络进行情感词分类训练,从而得到动态情感词典.使用动态情感词典进行初级扩展也可以得到更大的静态情感词典.实验结果表明使用该动态情感词典进行微博情感分析可以有效地提高分类精度.
Text sentiment analysis is an important branch of Natural Language Processing,and it is also a hot topic in current research.The most basic and crucial part of emotion analysis is to build a high quality emotion dictionary. In order to overcome the shortcomings of the traditional affective lexicons,such as emotional tendenability,this paper constructs a dynamic affective dictionary based on the structure of three layers of neural network. The first layer extracts the features of emotional information through the improved CBOW neural network; the second layer uses a bidirectional LSTMneural network to extract the two fork semantic dependency path features by using the two forked semantic dependency structure model,and the third layer,the current word features includes the center word information,the distance of vocabulary to the center word,the emotional features and semantic features obtained in the first two layers,which is the input for the bidirectional LSTM. Then,the bidirectional LSTMneural network is trained to classify emotional words,and a dynamic emotion dictionary is obtained. Using dynamic emotional lexicon for primary expansion can also get a larger static emotion dictionary. The experimental results showthat using the dynamic sentiment dictionary for micro-blog sentiment analysis can effectively improve the classification accuracy.
Text sentiment analysis is an important branch of Natural Language Processing,and it is also a hot topic in current research.The most basic and crucial part of emotion analysis is to build a high quality emotion dictionary. In order to overcome the shortcomings of the traditional affective lexicons,such as emotional tendenability,this paper constructs a dynamic affective dictionary based on the structure of three layers of neural network. The first layer extracts the features of emotional information through the improved CBOW neural network; the second layer uses a bidirectional LSTMneural network to extract the two fork semantic dependency path features by using the two forked semantic dependency structure model,and the third layer,the current word features includes the center word information,the distance of vocabulary to the center word,the emotional features and semantic features obtained in the first two layers,which is the input for the bidirectional LSTM. Then,the bidirectional LSTMneural network is trained to classify emotional words,and a dynamic emotion dictionary is obtained. Using dynamic emotional lexicon for primary expansion can also get a larger static emotion dictionary. The experimental results showthat using the dynamic sentiment dictionary for micro-blog sentiment analysis can effectively improve the classification accuracy.
引文
[1]Zhao Yan-yan,Qin Bing,Liu Ting,et al.Sentiment analysis[J].Journal of Softw are,2010,21(8):1834-1848.
[2]Song Jia-ying,He Yu,Fu Guo-hong,et al.Automatic expansion of domain-specific sentiment lexicon for Chinese[J].Journal of Chinese Information Processing,2015,29(6):75-82.
[3]Du Wei-fu,Tan Song-bo,Yun Xiao-chun,et al.A new method to compute semantic orientation[J].Journal of Computer Research and Development,2009,46(10):1713-1720.
[4]Teng Z,Vo D T,Zhang Y.Context-sensitive lexicon features for neural sentiment analysis[C].Proceedings of the 2016 Conference on Empirical M ethods in Natural Language Processing,2016:1629-1638.
[5]Xi Ya-hui.Construction of domain-specific sentiment lexicon in product review s[J].Journal of Chinese Information Processing,2016,30(5):136-144.
[6]Zhao Yan-yan,Qin Bing,Shi Qiu-hui,et al.Large-scale sentiment lexicon collection and application in sentiment classification[J].Journal of Chinese Information Processing,2017,31(2):187-193.
[7]Mohammad S M,Turney P D.Crowdsourcing a word-emotion association lexicon[J].Computational Intelligence,2013,29(3):436-465.
[8]Duyu Tang,Furu Wei,Bing Qin,et al.Building large-scale twitterspecific sentiment lexicon:a representation learning approach[C].Proceedings of COLING 2014,the 25th International Conference on Computational Linguistics:Technical Papers,2014:172-182.
[9]Yin Wen-ke,Zhu Ming,Chen Tian-hao.Domain thesaurus construction based on w iki hyperlink structure graph clustering[J].Journal of Chinese Computer Systems,2014,35(6):1286-1292.
[10]Graves A.Supervised sequence labelling with recurrent neural netw orks[M].Germany,Springer,2012.
[11]Hochreiter S,Schmidhuber J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[12]Kiperwasser E,Goldberg Y.Simple and accurate dependency parsing using bidirectional LSTM feature representations[J].Transactions of the Association for Computational Linguistics,2016,4:313-327.
[13]Cierniak R.A new approach to image reconstruction from projections using a recurrent neural netw ork[J].International Journal of Applied M athematics&Computer Science,2008,18(2):147-157.
[14]Qian Zhong,Li Pei-feng,Zhou Guo-dong,et al.Speculation and negation scope detection via bidirectional LSTM neural netw orks[J].Journal of Software,2018,29(8):1-21.
[15]Mikolov T,Yih W T,Zweig G.Linguistic regularities in continuous space w ord representations[J].The 2013 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies,2013:746-751.
[16]Tomas Mikolov,Ilya Sutskever,Kai Chen,et al.Distributed representations of w ords and phrases and their compositionality[C].Proceedings of Neural Information Processing Systems,2013:3111-3119.
[17]Tomas Mikolov,Kai Chen,Greg Corrado,et al.Efficient estimation of w ord representations in vector space[C].Proceedings of Workshop at International Conference on Learning Representations,2013:1-12.
[18]Chelba C.Exploiting syntactic structure for natural language modeling[D].The Johns Hopkins University,2000:225-231.
[19]Fancellu F,Lopez A,Webber B.Neural networks for negation scope detection[C].M eeting of the Association for Computational Linguistics,2016:495-504.
[20]Zhou P,Shi W,Tian J,et al.Attention-based bidirectional long short-term memory netw orks for relation classification[C].M eeting of the Association for Computational Linguistics,2016:207-212.
[1]赵妍妍,秦兵,刘挺,等.文本情感分析[J].软件学报,2010,21(8):1834-1848.
[2]宋佳颖,贺宇,付国宏,等.领域相关的汉语情感词典扩展[J].中文信息学报,2015,29(6):75-82.
[3]杜伟夫,谭松波,云晓春,等.一种新的情感词汇语义倾向计算方法[J].计算机研究与发展,2009,46(10):1713-1720.
[5]郗亚辉.产品评论中领域情感词典的构建[J].中文信息学报,2016,30(5):136-144.
[6]赵妍妍,秦兵,石秋慧,等.大规模情感词典的构建及其在情感分类中的应用[J].中文信息学报,2017,31(2):187-193.
[9]尹文科,朱明,陈天昊.基于Wiki链接结构图聚类的领域词典构建方法[J].小型微型计算机系统,2014,35(6):1286-1292.
[14]钱忠,李培峰,周国栋,等.基于双向LSTM网络的不确定和否定作用范围识别[J].软件学报,2018,29(8):1-21.
[2]Song Jia-ying,He Yu,Fu Guo-hong,et al.Automatic expansion of domain-specific sentiment lexicon for Chinese[J].Journal of Chinese Information Processing,2015,29(6):75-82.
[3]Du Wei-fu,Tan Song-bo,Yun Xiao-chun,et al.A new method to compute semantic orientation[J].Journal of Computer Research and Development,2009,46(10):1713-1720.
[4]Teng Z,Vo D T,Zhang Y.Context-sensitive lexicon features for neural sentiment analysis[C].Proceedings of the 2016 Conference on Empirical M ethods in Natural Language Processing,2016:1629-1638.
[5]Xi Ya-hui.Construction of domain-specific sentiment lexicon in product review s[J].Journal of Chinese Information Processing,2016,30(5):136-144.
[6]Zhao Yan-yan,Qin Bing,Shi Qiu-hui,et al.Large-scale sentiment lexicon collection and application in sentiment classification[J].Journal of Chinese Information Processing,2017,31(2):187-193.
[7]Mohammad S M,Turney P D.Crowdsourcing a word-emotion association lexicon[J].Computational Intelligence,2013,29(3):436-465.
[8]Duyu Tang,Furu Wei,Bing Qin,et al.Building large-scale twitterspecific sentiment lexicon:a representation learning approach[C].Proceedings of COLING 2014,the 25th International Conference on Computational Linguistics:Technical Papers,2014:172-182.
[9]Yin Wen-ke,Zhu Ming,Chen Tian-hao.Domain thesaurus construction based on w iki hyperlink structure graph clustering[J].Journal of Chinese Computer Systems,2014,35(6):1286-1292.
[10]Graves A.Supervised sequence labelling with recurrent neural netw orks[M].Germany,Springer,2012.
[11]Hochreiter S,Schmidhuber J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[12]Kiperwasser E,Goldberg Y.Simple and accurate dependency parsing using bidirectional LSTM feature representations[J].Transactions of the Association for Computational Linguistics,2016,4:313-327.
[13]Cierniak R.A new approach to image reconstruction from projections using a recurrent neural netw ork[J].International Journal of Applied M athematics&Computer Science,2008,18(2):147-157.
[14]Qian Zhong,Li Pei-feng,Zhou Guo-dong,et al.Speculation and negation scope detection via bidirectional LSTM neural netw orks[J].Journal of Software,2018,29(8):1-21.
[15]Mikolov T,Yih W T,Zweig G.Linguistic regularities in continuous space w ord representations[J].The 2013 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies,2013:746-751.
[16]Tomas Mikolov,Ilya Sutskever,Kai Chen,et al.Distributed representations of w ords and phrases and their compositionality[C].Proceedings of Neural Information Processing Systems,2013:3111-3119.
[17]Tomas Mikolov,Kai Chen,Greg Corrado,et al.Efficient estimation of w ord representations in vector space[C].Proceedings of Workshop at International Conference on Learning Representations,2013:1-12.
[18]Chelba C.Exploiting syntactic structure for natural language modeling[D].The Johns Hopkins University,2000:225-231.
[19]Fancellu F,Lopez A,Webber B.Neural networks for negation scope detection[C].M eeting of the Association for Computational Linguistics,2016:495-504.
[20]Zhou P,Shi W,Tian J,et al.Attention-based bidirectional long short-term memory netw orks for relation classification[C].M eeting of the Association for Computational Linguistics,2016:207-212.
[1]赵妍妍,秦兵,刘挺,等.文本情感分析[J].软件学报,2010,21(8):1834-1848.
[2]宋佳颖,贺宇,付国宏,等.领域相关的汉语情感词典扩展[J].中文信息学报,2015,29(6):75-82.
[3]杜伟夫,谭松波,云晓春,等.一种新的情感词汇语义倾向计算方法[J].计算机研究与发展,2009,46(10):1713-1720.
[5]郗亚辉.产品评论中领域情感词典的构建[J].中文信息学报,2016,30(5):136-144.
[6]赵妍妍,秦兵,石秋慧,等.大规模情感词典的构建及其在情感分类中的应用[J].中文信息学报,2017,31(2):187-193.
[9]尹文科,朱明,陈天昊.基于Wiki链接结构图聚类的领域词典构建方法[J].小型微型计算机系统,2014,35(6):1286-1292.
[14]钱忠,李培峰,周国栋,等.基于双向LSTM网络的不确定和否定作用范围识别[J].软件学报,2018,29(8):1-21.