Analysis And Application Of Chinese Sentiment Orientation In Financial Field

On the one hand,the domestic network utilization of the younger generation increases year by year,on the other hand,domestic,younger people tend to retail investors.Tracking and analyzing people in each big financial website,which posts the sentiment mood of public opinion monitoring is of great significance for the entire market.This thesis uses HowNet dictionary,Harbin institute of technology and other public feelings,then builds the basic sentiment lexicon.Combined with financial sector work experience,this thesis summarizes the financial sector sentiment dictionary.At the same time,in order to obtain corpora in the financial domain,this thesis designs and develops own orientation crawlers on Internet,used for directional climbing the shares post from the eastmoney website,as the original corpora.To calculate the SO_PMI value of words,this thesis uses way of mutual information with manual annotation corpora.Then combined with the artificial judgment threshold settings,it expands the basic sentiment dictionary and improves analysis accuracy of the follow-up model.Through the use of dependency syntactic relationships combined with sentiment dictionary,this thesis designs the model of sentiment analysis based on dependency syntactic and calculates the dependency relationships of groups,sentences and discourse as well as the sentiment intensity of posts respectively.After experiment on 6000 test texts,the prediction average accuracy rate reaches 69%.To compare the analysis accuracy of different classification methods,this thesis trains and classifies the test language material bases on the word vector Naive Bayes and C4.5 decision tree classification methods respectively.The corresponding average accuracy rate of both are no more than 60%.This fact initially verifies the validity of classification model in this thesis.Finally,this thesis synthetically applies the above work content for designing and developing sentiment based financial analysis system as well as realizing data acquisition and the sentiment polarity analysis process of stock.