Experiments with Semantic Enrichment for Event Classification in Tweets

Twitter has become key for bringing awareness about real-world events, but the identification of event related posts goes beyond filtering keywords. Semantic enrichment using knowledge sources such as the Linked Open Data (LOD) cloud, has been proposed to deal with the poor textual contents of tweets for event classification. However, each work considers a particular type of event, underlined by specific assumptions according to the application purpose. In a search for an approach that suits different types of events, in this paper we identify different types of semantic features, and propose a process for semantic enrichment that involves the mapping of textual tokens into semantic concepts, the extraction of corresponding semantic properties from the LOD cloud, and their interpolation for event classification. We evaluate the contribution of each type of semantic feature using different tweet datasets representing events of distinct natures, and knowledge extracted from DBPedia.