Abstract

Tweets contain mentions of numerous entities, persons and events, and often additional information, like an opinion, that can be viewed as an annotation of that entity. However, this information is currently being accumulated only by spe- cific applications without being made available in a generic format. We discuss a natural language processing approach to extract information about entities and their annotations from tweets and transform them into a semantic, reusable knowledge base. We believe this will greatly facilitate access to user-generated Twitter data for many applications.