Research On User Sentiment Analysis For Social Media Networks

In recent years,the rapid development of Internet,communication between people is becoming more and more convenient and quick,ever-bubbling spring,with the new multimedia social platform around the growing social platform of communication,every day hundreds of thousands of users through the social platform of their comments and views.This text information contains a large number of emotional tendency(comment on something "good" and "bad")words,these words with emotional tendencies can reflect user emotional state at that time.On the one hand,the multimedia social platform has evolved into the comment text database of opinion mining and sentiment analysis.On the other hand,the development of the multimedia social platform also raises the higher demand for the mining and emotional analysis of opinions.Film comment text is relatively common comment on information,multimedia networking platform for movie reviews research has a lot of very perfect,in the traditional comments in the field of sentiment analysis is bad in the breakthrough,so this article on the basis of the comment based on the traditional information combined with some other factors to the comment text sentiment analysis,users can be found through the analysis of the emotional movie features,and based on the characteristic information of the film,deduce the user's preference,this paper presents a moving film the implied meaning of the text information content and score combined to build a new movie recommendation model.So this article is based on multimedia social platform(movie reviews BBS)comment text information in order to develop the film recommendations and grading forecast method were studied.And to conduct the following three aspects:1.For movie review the words in the text information statistics and analysis,based on hownet dictionary built a emotional lexicon in the field of film,used to emotional movie review information classification,according to the classification of emotional spread theme in this feature,in this paper,a classification method based on emotional words extraction of movie reviews.Using machine-learning methods to categorize text,it gets its emotional bias.In the primary emotion classification is proposed on the basis of the combination of statistical BSI secondary emotion classification model,secondary emotion classification model has been proved by the experiment contrastlevel of emotion in comparison with the accuracy of classification increased by 5%,and have not marked the comment text has better outstanding emotion classification ability.2.Users when browsing the web or search content to have the existence of the browsing history,the log data with the increase in the number of the operation of the user,user log data can reflect some habit or preference,the log data statistical analysis,in order to get the user's search behavior characteristics,can undertake related behavior characteristics of the user,thus can get user search behavior of history,these users search behavior of history for user ratings increase has played a certain influence to the precision.Due to the large number of logs,this article is based on the Hadoop framework.Put forward a new method of recommended to review the content and user ratings of the implied,the user log information combination of statistical data,we design a new movie recommendation model,the first use of topic themes distribution in the text mining,subject distribution is then used to depict the user's preferences and film portrait,combined with the user's history of statistics and BSI emotion to build a score prediction model,and then based on logistic regression model on the training topics and the inner link between the user's score,the algorithm enriches the recommended data information,can effectively alleviate when user cold start,recommend scale sparse problem in the system.Then,using the real data to compare the experiment,the experimental data were used to validate the model.