Topic Trend Detection in Newsgroups

In this paper we report on a project for Corporate Technology of Siemens AG, Munich, which aimed at the use of a social network analysis approach for trend and trend shift detection over time in a technically oriented newsgroup on mobile phones. The analysis was based on the assumption that a trend shift is relevant only if relevant (central) members of the newsgroup initiated this shift. A shift could occur in one of two ways. Either within a subgroup the topic shifted, e.g. other words were used, or the relevance of a group within the newsgroup shifted, and the now more relevant subgroup discussed a different topic. We used eigensystem analysis as a method and could show that as groups shifted over time, so did topics.