Abductive Requery

摘要

We present an abductive approach to query formulation based on ontology and semantic rule content. Traditional Decision Logic ontologies offer scant opportunity for the application of abductive inference. When augmented by semantic rule languages such as SPIN, however, abduction becomes an attractive possibility for the targeted exploration of potentially large or virtual data sources. We describe a method of pre-processing such rules for use in abductive reasoning. We also describe rule selection and re-query formulation methods once a set ofA-Box statements is presented to the abductive reasoner. Our approach assumes that the majority of A-Box content is virtual, ie, not realized until specifically asked for from non-ontology sources such as databases and/or unstructured text documents.