Font Size: a A A

Sentiment Classification Of Uyghur Text Based On Deep Learning

Posted on:2018-05-17Degree:MasterType:Thesis
Country:ChinaCandidate:S H WangFull Text:PDF
GTID:2348330533456162Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With the development of basic communications services in Xinjiang,a large number of social networking sites have been established.In these social networking sites there are a lot of subjective remarks,which contains the user's rich emotions.However,due to the rapid update of social networking sites,large amounts of data,the manual way to deal with these data will consume a lot of manpower and material costs,so it need to study how to use computer to automate the processing of these data.In view of this problem,the text emotion analysis came into being.The existing emotional analysis is mainly focused on English,Chinese and other languages,for similar language like Uygur research is relatively less.The traditional emotional classification method does not take into account the deep semantic relations among the texts,rely too much on the background knowledge and ignores the emotional polarity of the text.In this paper,we use the deep learning technique to extract the semantic features of the text and explore the Uygur language emotion analysis technique based on the deep learning.First of all,this article uses the word vector training tool to construct emotional dictionary.By obtaining the semantic representation of the word,the Chinese emotional thesaurus is used to construct the Uygur emotional dictionary by using the semantic similarity calculation.Compared with the traditional method of using experts in the field or a large number of corpus artificially to construct emotional dictionary,this method can save a lot of manpower and resources.Secondly,in order to solve the problem of poor generalization ability in traditional traditional research methods,this paper explores the Uygur language emotion analysis method based on word vector and LSTM-RNN model.First,the word vector training tool is used to obtain the word vector representation of Uygur text,and the deep semantic information between Uygur words,which is used to improve the expression of feature vector to text semantic information.Secondly,by exploring the advantages of LSTM-RNN neural network and taking into account the context information,it is better to obtain the text history information.Combined with the advantages of the two,complete the Uygur language emotional classification,and the experiment obtains better results.Finally,aiming at the problem that the simple word vector can not take into account the emotional information between texts,a Uygur language emotion analysis method based on bidirectional LSTM model is proposed.By constructing the emotional dictionary,this method combines the traditional way of extracting the vector features of the emotion with the word vector based on the language model to constitute the vector representation of the data.This method not only takes into account the semantic representation of the text,but also takes into account the emotional information between the texts.Then,the bidirectional LSTM model is used to obtain the historical information of the forward direction and the backward direction.Finally,it is shown that the method can fully describe the emotional information of the original text data,and then improve the classification accuracy of the model.
Keywords/Search Tags:Uighur, deep learning, long and short memory neural network, emotional analysis
PDF Full Text Request
Related items