The Research On Mongolian Micro-blog Sentiment Analysis Based On Deep Learning

With the development of social media,more and more users express their feelings through micro-blogs.Sentiment analysis and mining of a large number of texts on micro-blogs has great application value.In recent years,deep learning technology has been widely used in the field of NLP.Therefore,this thesis studies Mongolian micro-blog sentiment analysis based on deep learning,and the specific work contents are summarized as follows.(1)Due to the lack of Mongolian language data,this thesis constructed a Mongolian micro-blog sentiment analysis corpus.Firstly,we selected part of NLPCC2014 Chinese corpus as the original corpus.Then,a large amount of noise can be removed by preprocessing steps.And,the cleaned original corpus was translated by Chinese-Mongolian machine translation tools,and the translated Mongolian corpus were manually corrected.Finally,the Mongolian Sentiment Analysis Corpus was constructed.(2)Aiming at the subjective recognition task of Mongolian micro-blog,this thesis proposes a sentiment analysis model called MA-BLSTM,which used multi-head self-attention mechanism combined with BLSTM to extract features from texts.In the experiment part,the optimal super parameters of the model were found through relevant experiments,and the performance of the MA-BLSTM model was verified to be superior to other baseline models on the Mongolian micro-blog corpus.(3)To judge the sentiment polarity of Mongolian micro-blog,this thesis used hybrid model TCNN,which combined Transformer with CNN.In the experimental part,to achieve the optimal performance of our model,the number of encoders and the number and size of convolution kernel are determined by repeated experiments.As a result,it is verified that the effect of the TCNN model proposed in this thesis outperforms other baseline models on the sentiment analysis data set of Mongolian micro-blog.(4)Aiming at Mongolian micro blog texts,this thesis develops a B/S structure sentiment analysis system.The functions of the system include one-key analysis,subjective recognition and judgment of sentiment polarity.Meanwhile,in order to expand the data set,the corpus and results after each sentiment analysis are saved to facilitate the subsequent updating and iteration of the model.