Abstract

Information extraction systems have been successfully deployed for domains ranging from terrorist activities to medical records. However, building these systems remains costly for users who lack annotated training corpora or knowledge engineering expertise. This paper proposes a framework for an interactive information extraction environment in which the user trains the system by example and by feedback about performance. If successful, this will be the first system that allows end-users to create information extraction systems without the aid of computational linguists and NLP system designers.

Citations