Research Of Region Feature Discovery Based On Web Comments

Internet plays an important role in people's daily life and has become a significant platform for obtaining information,the development of Web2.0 technology changes the way that people communicate via Internet.People now not only just obtain information from Internet,but also participate in providing and spreading information.The public tend to express their views,opinions,take part in the topic discussion and interaction on the network platform.Many online news sites provide timely news and social hot news,etc.and they also provide users to participate the discussion and make comments including user's nickname,location and commenting time.Among them,the user's location information shows the user's life and activity venue,users in the same geographic location might have the similar comments behavior.Analyzing users comments behavior within a region,we can understand the implied regional feature and hot topics in a region.This paper proposes region feature discovery based on user comments behavior.Region feature or hot topics can be a category event or a topic.Region feature can be labeled by the specific topic or event that users comment on.Analyzing the region future can lable the region topic and can be used to advertise and perform some other business strategies.It also can provide reference for reion administration and planning which has practical significance.This paper proposes discovering region topic feature based on semantic and attention extent according to users' comments data of news.Semantics and attention extent exhibit users'comment behavior.The semantics exhibits users' comments and the attention extent exhibits the number of user's comment of news.Users' comments semantic information can express users prefer topic and the comments number can exhibit that how much users concern on the news.The previous related works are mostly based on the users' comment behavior initiatively.However,this paper analyzes region feature based on users comments behavior for the specific topic which can reflect users comment behavior better and discover the region-feature accurately.This paper first preprocesses text data of user comments in region feature discovery based on semantics.Preprocessing process includes Chinese words segmentation,stop words removal and region feature words extraction.Then modeling the semantic comments data that contains region feature phrases and learning the topic feature of each region through LDA algorithm.This paper also cluster regions based on the document-topic feature matrix to find the regions with similar topic feature.This paper proposes region feature and abnormal region query problem in region feature discovery based on attention extent,and proposes three region feature query algorithms and one abnormal region query algorithm which are maximum eigenvalue query,minimum eigenvalue query,maximum offset distance query,and the overall offset distance query method.This paper uses the real dataset of Sina news in experiment and select eight categories of news topic and 31 provinces in China.The experimental results show the regions' topic tendency and region clusters.This paper analyzes the query results and compares the three region feature query algorithms.The experimental results show the significance of the proposed problem.