Scalable Web Reasoning Using Logic Programming Techniques

One of the key issues for the uptake of the Semantic Web idea is the availability of reasoning techniques that are usable on a large scale and that offer rich modelling capabilities by providing comprehensive coverage of the OWL language. In this paper we present a scalable extension of our A Box reasoning framework called DLog. DLog performs query-driven execution whereby the terminological part of the description logic knowledge base is converted into a Logic Program and the assertional facts are accessed dynamically from a database. The problem of instance retrieval is reduced to a series of instance checks over a set of individuals containing all solutions for the query. Such a superset is calculated by using static-code analysis on the generated program. We identify two kinds of parallelism within DLog execution: (1) the instances in the superset can be independently checked in parallel and (2) a specific instance check can be executed in parallel by specialising well-established techniques from Logic Programming. Moreover, for efficiency reasons, we propose to use a specialised abstract machine rather than relying on the more generic WAM execution model. We describe the architecture of a distributed framework in which the above mentioned techniques are integrated. We compare our results to existing approaches.