Reference: McGuinness, D.L.; Pinheiro da Silva, P.; Murdock, J. W.; & Ferrucci, D. Exposing Extracted Knowledge Supporting Answers. Technical Report, Knowledge Systems, AI Laboratory, Stanford University., 2005.
Abstract: As more systems rely on knowledge bases built from automatic and semi-automatic methods, it is becoming more important to provide solutions that not only answer questions but also provide information describing how the answers were obtained. We aim to make answers more useful that rely on the combination of knowledge bases, some of which may have been built as the result of extraction processes and others may have been hand constructed from reliable sources. Our solution provides an infrastructure capable of encoding justifications for answers in a single format. This provides an end-to-end description of the knowledge derivation process including access to the raw text documents, descriptions of the text analytic processes used for extraction, as well descriptions of the ontologies and many kinds of information manipulation processes, including standard deduction. We have implemented our solution and we are using it in several sponsored projects.
Full paper available as pdf.