Mining burst topical keywords from microblog stream

Microblog is becoming more and more popular in daily life, through which people can exchange small elements of content, such as short sentences, individual images, or video links. Considering millions of data produced every day, a challenging issue is how to quickly know current burst topic from large volume of continuous microblog streams. This work proposed a method for mining burst topical keywords from microblog stream based on incremental calculation method and burst theory. Considering burst topic´s immanent characters, such as burst phenomenon and uncommon, a method combining these characters was proposed to weight the words of microblog stream in a fixed time interval (Time Window). Eventually, those words whose weight is over a given threshold are selected as burst topical keywords of that period. The experimental results show that the proposed method can mine the burst topical keywords which can correctly reflects the trends of burst topics in microblog.