Workshop on Knowledge Extraction and Consolidation from Social Media (KECSM 2012), Boston, USA, November 2012
Abstract. An increasing number of innovative applications use data from online social networks. In many cases data analysis tasks, like opinion mining processes, are applied on platforms such as Twitter, in order to discover what people think about various issues. In our view, selecting the proper data set is paramount for the analysis tasks to produce credible results. This direction, however, has not yet received a lot of attention. In this paper we propose and discuss in detail a platform for supporting processes such as opinion mining on Twitter data, with emphasis on the selection of the proper data set. The key point of our approach is the representation of term associations, user associations, and related attributes in a single model that also takes into account their evolution through time. This model enables flexible queries that combine complex conditions on time, terms, users, and their associations.