Research On The Extraction And Analysis Methods Of Chinese And Vietnamese Bilingual News Viewpoints

Vietnam is one of the important neighbors of our country.There is becoming increasingly close between China and Vietnam in many fields.It is important to analyze and master the public opinion of the news of both country.However,there are massive news texts on the Internet,and it is time consuming and laborious to analyze and summarize them manually.Therefore,it is of great significance and value to study the method of automatic analysis of Chinese and Vietnamese bilingual news texts.The news text mainly contains two parts:the description of the objective facts that have occurred and the subjective judgments of the objective facts.The vocabularies of the objective facts are the attributes of the news,such as "name,place name,institution name",subjective judgment is express by sentiment words,such as "meaning,influence,praise" and so on.Based on this,this paper fusion news element relation and the sentiment relation into the graph model,studies the method of opinion sentence extraction based on the graph model.On this basis,studies the method of summarize the difference of opinion sentence,and make sentiment classification.This paper mainly completes the following characteristics research work:1.Opinion sentence extract of Chinese and Vietnamese news based on attributes and sentiment relationNews contains attributes and sentiment words regardless of the language,combined with this feature,we proposed a method to extract opinion sentence based on attributes and sentiment relation.Firstly,build a sentence relation graph model according to the attributes and sentiment information of the sentence.Then,the weights of the edges in the graph model are calculated according to the attribute relation strength and sentiment relation strength,and the graph model is solved to relatize the extract of opinion sentence.2.Summarize the difference of opinion sentence based on graph modelFor the same news events,Vietnamese news and Chinese news held different views.In order to extract the difference between Chinese and Vietnamese bilingual news,on the basis of the first research work,a method of summarize the different of opinion based on undirected graph model is proposed.This method build a bridge between different languages based on machine translation.Firstly,the similarity of Chinese and Vietnamese sentences are calculated,according to the similarity to make a filter.Then,a graph model is established,the edge of the same language is similarity,and the edge between difference language is difference degree.Finally,according to the above two similarities,the random walk algorithm is used to extract the difference between different languages.3.Sentiment classfication based on convolution neural networkOn the basis of extracting the difference viewpoints,in order to further analyze the sentiment polarity of the opinion sentence,a method of sentiment classification based on convolution neural network is proposed.Compared with the traditional method,this method does not need to construct sentiment dictionary and carry out the complicated feature extraction work.The key to solving bilingual sentiment classification using convolution neural networks is how to map the characteristics of different languages to the same space.In order to solve this problem,first,to collect a large number of Chinese and Vietnamese not labeled corpus,respectively,training Chinese word vector and Vietnamese word vector.Then,for Chinese sentences,use machine translation to map it to Vietnamese,the same,for Vietnamese sentences,using machine translation to map it to Chinese.Finally,the sentence vector vector and the Vietnamese word vector are trained as different channel input convolution neural network models to realize the judgment of sentiment polarity.