Emotional Tendency Analysis Of Uyghur Text Based On Deep Learning

Emotional analysis is one of the important topics in the field of natural language processing.The main purpose is to make valuable information prediction based on mining results.With the overall improvement of the level of Internet development,Internet users in XinJiang increased significantly,the Uyghur language network communication platform is also established.The text with its own subjective feelings,reflecting the attitude,position and opinion of the Uyghur commentary is increased dramatically.Therefore,This paper focuses on the analysis and research of Uyghur commentary texts rich in content and large amount of information.Most of the existing methods of emotional analysis are based on the traditional machine learning model or the combination of artificial extraction characteristics,to enhance system performance,but there are some limitations: 1)the limitation relying on artificial experience to extract the characteristics of the sample rules.2)Base on traditional machine learning,the generalization of complex problems is limited,even can not fully express deep-level semantic information.Therefore,for the Uyghur emotional analysis task,this paper completed the following work(1)We propose a emotion analysis method based on deep semantics of Uyghur using depth learning algorithm.According to the specific linguistic characteristics of Uygur language.The influence of the part of speech,emotional words,turning conjunctions and negative components on the emotional tendencies is considered,and introduce word embedding is full of semantic and contextual information,using the unsupervised stack auto-encoder learning deep semantic features.(2)It is prerequisite for the correct analysis of sentence-level emotional that reference items are accurately removed ambiguity.This paper analyses the present situation of the research on the anaphora resolution in recent years,and puts forward the use of the deep learning method to complete the anaphora resolution task of Uyghur noun phrase based on deep semantic information.According to the specific linguistic characteristics of clanguage,five kinds of noun phrases with referentiality are summarized and the corresponding 13 features are extracted.Using word embedding to express lexical features,and increase the expression of text semantics and syntactic information.(3)Considering the influence of word order and noun phrase context information on the resolution results,we propose a multi-model approach.The method introduce two-layers LSTM learning noun phrase contextual semantic information,mining implicit relationship between noun phrases.In order to further extra the text semantics,The multi-layer mapping unit of the deep stack auto-encoder is used to further study the deep semantic and syntactic information.