Mining Newsworthy Events in the Traffic Accident Domain from Chinese Microblog

Abstract

Microblog can provide a valuable resource for journalists as it captures potential newsworthy events as they occur, including ones occurring remotely. Given the large volume and the fast pace of typical microblog, it is impractical to monitor all microblog postings for potential news events. Therefore, it would be useful if a method exists that uses text mining to help identify such events. For this endeavor, we need a good model of newsworthiness that furthermore can be operationalized with text-mining techniques. This study examines the feasibility and usefulness of such a model by first adopting the Shoemaker model of newsworthiness, one of the most comprehensive and accepted among such models; refining it based on a set of extensive interviews with domain experts and users in the context of news media in China; operationalizing it with a set of text-analytic measures in the domain of traffic accident; and testing its feasibility and validity using data from Weibo, the largest microblog site in China. As such, we believe that this study makes important theoretical and methodological contributions by developing and testing the most comprehensive and computable model of newsworthiness to date. We also point out its limitations and the areas that need further research.