A framework for detecting contextual concepts in texts
Although information extraction systems often focus on assertions about conrete entities and relationships among them, there is growing interest in tools for identifying the context in which these assertions are made. Important elements of context include the source of the information, the author s attitude towards the informationor his evaluation. The difficulty of automatic processing lies in the fact that usually there are no conventional ways, or set expressions for conveying concepts of contextual elements. This leads to a great variety of surface forms that traditional information access systems do not detect. We propose a framework to extract contextual concepts. We present its basic principles and properties as well as challenges for future work.
Electra Workshop, Pestana Bahia, Salvador, Brazil, August 15-19, 2005.