Micro-blog Topic Detection Based On Users' Interests And Communities

With the rapid development of Internet technology,Network media has become necessary in people's daily life.Weibo,as one of the typical emerging,as a breakneck speed affects the pattern of social communication.Users can through the microblogging anywhere at any time to update his or her personal status and participate in discussions or concern the topic he loves,it makes microblogging becomes the gethering place of hot topics.Microblog topic detection not only can offer hot topic infomation for users,but also support for government in Emergency monitoring and public opinion analysis,so it has a very important practical significance for Study how to detect the hot topic in mass microblogging information.The microblog text has a large gap with traditional text.It not only has a large number of ellipsis,refer to langauge and subjectivity of personalized language,but also has the for Short text,strong topic discreteness,real-time and interactive features.So the traditional topic detection method is not apply for microblog topic detection,in this paper,combine with the characteristic of microblog,put forward a set of microblog topic detection method based on user interest and community structure.First,we bring in the concept of user community in the topic detection algorithm,quantitative the attention of users to relation network for microblog user,next using community discovery algorithm to identify the community in this complex network and add interest tags for community,and detect topics in community interval.Aim at the defects of sparse feature and strong noise,we put forward a microblog topic detection algoritlhm combined with word importance and ? neighbor graph,the algorithm takes the word important degree as the starting point,extract max k words as key word set,and calculate the similarity between microblogs,get microblog cluster using graph cut algorithm,fimally,get the theme word based on key importance.In real microblog data set on the experimental results show that the algorithm can significantly remove noise,detect hot topics in community quickly and accurately.Aim at the character of heat,occurence and time feature for topics,we put forward a microblog topic detection algorithm combined with word importance and time window.The algorithm firstly divide microblog text based on time window,and get the word importance based on word heat feature and occurence feature in time window interval,extrack max k words as the candidate theme words.Next,building an term co-occurrence network using candidate theme word,and discovery community in this erm co-occurrence network with appropriate community partition algorithm,finally,we get the topic and ranking based on word importance.In real microblog data set on the experimental results show that the algorithm can detect topics in time window interval quickly and efficiently,avaluate topic importance feature,heat feature and occurence feature,can do real-time tracking for Topic importance changes over time.