Encyclopedia Fact Extraction, Querying and Answer Generation: Based On MRS Representations of Natural Language

Abstract

This master's thesis explores the possibility of using the Linguistic Knowledge Builder (LKB) in conjunction with a Prolog based program in a framework which integrates encyclypedia search for finding answer to natural language questions. I outline an architecture and a proof-of-concept implementation that shows that this is possible. With the current implementation, answers can be found when the question is just an anaphoric or simple ontological relational reference away from the sentences in the encyclopedia article. The modular design of the framework will allow incremental future improvements in all areas from fetching, reasoning about and answering questions for encyclopedia articles.