Obtaining Formal Knowledge from Informal Text Analysis

Populating formal knowledge-bases from natural-language text is a long-standing objective in computer science. Recent advancements in both ontology research and information extraction research are making this objective increasingly obtainable. However, there are still serious obstacles to performing automated reasoning over the contents of text documents. This paper focuses on one of those obstacles: differences between the formal ontologies used by reasoning systems and the informal ontologies used by extraction systems. We describe a framework for automating translation from extracted information to formal knowledge, and we describe a complex, implemented system that uses this framework. We also describe results from this system applied to a moderately large (approximately 75 MB) text corpus.