Research And Implementation Of Network Text Sentiment Analysis

With the rapid development of the Internet,our society has entered a new information age.More and more people has become the user of Internet,they are willing to express their real emotion in the network and to share their personal opinions.We can accurately express our emotional with positive or negative attitude to product evaluation,real-time hot news commentary and business service evaluation.It is benefit to improve the development of business product and regulatory work of government from extract the evaluate information.With the coming of the era of big data,we can't use traditional manual methods to extract and analysis the meaningful comment information with the rapidly growth of information.Therefore,the research of sentiment analysis to network text has great significance.The main purpose of text sentiment analysis is to mine the emotional information and recognize the emotion polarity in the text.The main research of this thesis is as follow:(1)A kind of feature extraction method of lexical sentiment is proposed.This paper finds all lexical features with emotional information from sentiment characteristics corpus based on lexical rules in the first.And then it is extract the lexical emotional characteristics by the maximum matching algorithm of lexical,which based on accuracy and occupancy of sentiment recognition.Finally compare with the proposed and existing method through experiment.The experiments prove that this method of feature extraction based on sentiment dictionary can effectively enhance accuracy.(2)A kind of sentiment analysis method which combing the sentiment dictionary and SVM is proposed.Firstly,this paper obtains part of highly accuracy classification results based on dictionary analysis algorithm.Then the classification results used to train the SVM classification,which would classify the rest of text.It can get the last result from combination with lexical feature and SVM in the last.Finally,it is to make an experiment in three fields about product,hot news and film with proposed method and several other existing methods.The experiments show that the proposed method has better judgment results than the methods of dictionary-based and machine learning.Besides,the proposed method has better effect than the existing methods of combing the sentiment dictionary and machine learning in recognition accuracy.(3)This paper design system of sentiment analysis about network text based on proposed method and existing classification method.The system contains data acquisition,data processing and data analysis function.It can quickly to analysis online network text.The system has great application value.