5WTAG: Detecting the Topics of Chinese Microblogs Based on 5W Model

Abstract

A hash tag is an important metadata in micro blogs and used to mark topics or index messages. However, statistics show hash tags are absent from most of the micro blogs. It poses great challenges to the retrieve and analysis of these tagless micro blogs. In this paper, we summarize the similarity between micro blogs and short message news, and then propose an algorithm named 5WTAG for detecting micro blog topics based on 5W (When, Where, Who, What, how) model. Since 5W attributes are the core components in event description, it is guaranteed theoretically that 5WTAG can extract the semantics of the micro blogs properly. We introduce the detailed procedure of 5WTAG in this paper including the candidate hash tag construction and recommendation computation. Finally, we verify the semantical correctness of the candidate hash tags as well as the effectiveness of recommendation computation using the real data set from Sina Weibo.