Abductive retrieval of structured documents

In this paper, we propose to employ abductive reasoning in logic-based information retrieval in the field of structured document handling. Following van Rijsbergenʹs suggestion, we regard retrieval as an inferential process that determines those documents that logically imply a query. Abductive reasoning allows the design, processing, and development of a domain model bottom-up with respect to the underlying document structure in a straightforward manner. We describe the path from abductive inference to an information retrieval system. Our prototype system is demonstrated by tracing an example of building and querying a semantic model derived from an SGML-structured document collection on art and artists.