Research On Topic Tracking And Tendency Analysis Techniques For Web Texts

With the popularization and rapid development of the network,massive network data has been generated.People achieve information through Web texts such as portal news,micro-blog,BBS and other network texts.Topic tracking can help people to explore the process of topic development from the huge and complex network texts information.The research of sentiment analysis can get the emotional state people want to express through the text.Topic tracking and the sentiment analysis provide data support for public opinion monitoring,and are important research directions in data mining.This paper focuses on topic tracking and sentiment tendency analysis of the texts.The main work is as follows.1)Combining with the characteristics of online news texts,topic models have been selected for topic modeling and tracking analysis.The network news texts are divided according to the time window,and the probability distribution of the documents topic is obtained by using the LDA model,and the parameters of the model are inferred by the Gibbs Sampling method.And then the improved Single-Pass method is utilized to track and analyze the topic,and JS divergence is used to measure the distance of topic.The time factor is introduced into the similarity calculation to improve the similarity of time closely topics.The topics are processed in batches according to the time window,and then are classified by the similarity size,so the problem that the input sequence influences the tracking results is solved.Finally,the content development and topic intensity changes are analyzed in different time windows.2)The method based on word vectors is proposed to extend the sentimental features.The word vector is obtained through obtaining the network news.The comprehensive selection of potential emotional words is studied to expand the emotional vocabulary based on the distance between words and the basic sentimental words.This method can cross-domain and quickly obtain potential sentimental words.After the domain expert selection,extended emotional words can be used as basic emotional words for sentimental analysis.3)The sentiment blending model based on affective features is used to study the tendency of news texts.The extended sentiment words can be applied to the topic-sentiment mixed model for the textual sentiment analysis.The topic related sentimental tendency is analyzed through experiments,which is a reference for the study of text topics and sentimental trends.