Inferring dynamic taxonomies for terms based on UGC

Abstract

Users of the web are currently the main content authors. Social networks form a valuable source of large volumes of user-generated content that might be beneficial in different areas of research. In this paper, we exploit the data generated by users of micro-blogging services, in particular Twitter, for disambiguating terms and inferring possible taxonomies for terms. We conducted an exploratory study to test the possibility of inferring high-level categories and possible sub-categories for which terms might be included. This experiment exploits the collection of hashtags which are mentioned with a specific hashtag in Tweets text., and the Open Directory Project (ODP), in order to discover dynamic taxonomies for a term (hashtag) with no knowledge needed about the semantic meaning of that term. We present several experiments which we have used to test this method, and promising results are reported.