Research On Mongolian News Text Classification And Sentiment Tendency Analysis Based On Deep Learning

In the current era of information explosion,it is becoming more and more difficult to obtain the required information,and the news media is an important means of obtaining information.News text classification and tendency analysis can be applied to news clue collection,public opinion monitoring,decision-making assistance,harmful information filtering,operation analysis,special topic analysis,news push and other fields,which can effectively help users obtain the required information quickly and accurately,and help the news media analyze the rules and related items in the vast data,so as to achieve the purpose of processing news conveniently and swiftly.At present,text classification and sentiment tendency analysis are studied more at home and abroad,and the development is comparatively mature,but the Mongolian-related research has a relatively late start and few results.With the development of Mongolian information technology,the content of Mongolian online news is gradually increasing.The classification of Mongolian news text and the analysis of emotional tendencies have become important topics.The research goal of this paper is to classify Mongolian news text and analyze its emotional tendency based on deep learning methods,and design and implement a Mongolian news text visualization system based on the research results.Firstly,the Mongolian news text data set is constructed through crawling,translation,preprocessing and other processes,and word2 vec model is used for word embedding training and feature representation to improve the accuracy of the model.Secondly,this article makes a comparative study on the classification models of Mongolian news texts.The feature vectors pre-trained by word2 vec are input into LSTM,Bi LSTM,W-LSTM and W-Bi LSTM models respectively to carry out a comparative experiment of Mongolian news text classification.Finally,w-bilstm model with the best classification effect is selected as the baseline for the Mongolian news text classification model and the Mongolian news text tendency analysis task.Thirdly,a model for analyzing the tendency of Mongolian news is constructed.This paper divides the emotional polarity of news text into positive,negative and neutral.By constructing a model of Bi-LSTM+Attention mechanism,tendency analysis is carried out.The application of Attention mechanism in news sentiment analysis can not only improve the performance of deep learning model,but also visualize the Attention score of the model,and intuitively see which parts of the whole document are given a higher weight by the model.Finally,a visualization system for Mongolian news text is realized.This paper designs a news text visualization system,using the above-mentioned training model based on deep learning and word2 vec statistical functions,through Flask+pyecharts,it realizes the Mongolian text including histogram,radar chart,map,pie chart,Attention visualization,etc.The news text is analyzed in multiple dimensions,and the judgment result and judgment basis are displayed.