Study On The Topic Discovery For Internet Group Chat Based On User Interactive Behavior

As the initial form of the cyber group,the Internet group consists of Internet users with same value and interests.Different with one-to-one and one-to-many relationship models,the interaction relationships in Internet group belong to many-to-many.The Internet group is complicatedly structured like a network with irregularly distributed nodes.When users discuss in the Internet group,a large number of conversational texts which contain valuable intelligence will be generated.The texts belonging to interactive short texts are characterized in that: colloquial expression,multi-role interaction and multi-topic interweaving.Therefore,traditional topic extraction methods are not fully applicable for the Internet group topic discovery research.Based on this,this paper attempts to propose a topic discovery method for the Internet group based on users' interactive behaviors.This method combines users' interaction relationships recognition with topic extraction.On the one hand,we can explore the social network structure of the Internet group,which offers a reference for users' preference judgments and users' relationship maintenance.On the other hand,we can provide a new idea for topic extraction methods to improve the effect of topic extraction.Firstly,this paper summarizes relevant research on topic discovery for socialized short text and interactive text,thus comprehend the achievements and shortcomings of the existing research.Secondly,this paper reviews the concepts and types of users' interactive behaviors and summarizes the related studies on users' interactive behaviors.Then it proposes a topic discovery method based on users' interactive behaviors,which mainly includes users' interactive behaviors recognition,users' interaction network construction,interaction intensity measurement,key users identification,user community detection,text integration and topic discovery.Finally,we take QQ group "Tuan Ren Tang" as an example to carry out our experiments.The results showed that:(1)There are explicit and implicit interactive behaviors in Internet groups,which correspond to explicit and implicit interaction relationships.(2)The interaction intensity is affected by four dimensions: amount of time,emotional intensity,intimacy and reciprocal services.(3)Users' interaction network includes leaders,core participants,active users and general users.The whole interaction network can be divided into three types of communities: intensive community,central community and sparse community.(4)The topics of group chat involve library service,library development,library resources,literature reading,and teacher-student relationships.The contents are all related to the library and information science.Then,we find that the user's interests are related to their identity,gender and work environment.Meanwhile,we find the topics of the central communities are controlled by the central users and key users deeply discuss topics in a wide range.