Ontology Query Answering on Databases

A paper written by Yue Pan, Li Ma and Jing Mei. It was presented at the ISWC2006.

[edit] Abstract

With the fast development of Semantic Web, more and more RDF and OWL ontologies are created and shared. The effective management, such as storage, inference and query, of these ontologies on databases gains increasing attentions. This paper addresses ontology query answering by means of a Datalog program, specifically tailored to bridge ontologies and databases. Introducing meta integrity constraints inspired by epistemic interpretations, we believe such a Datalog program suitable to capture ontologies in the DB favor, while keeping reasoning tractable -- Here, we present a logical equivalent knowledge base whose (sound and complete) inference system appears to a Datalog program. As such, a deductive RDF database is responsible for SPARQL query answering involved in OWL and SWRL. Bi-directional strategies, taking advantage of both forward and backward chaining, are then studied to support for this kind of customized Datalog programs, returning exactly answers to the query with respect to its logical framework.

The schedule for this talk can be found in the conference programme and a linked list of all talks is provided in the article on ISWC2006 papers. This article has originally been created from the RDF metadata for ISWC 2006.

[edit] Discussion

The following was moved from the discussion page where it was entered by someone attending the paper presentation. It should not be considered as an undisputed fact, but rather as a point that might require clarification by the authors.

[edit] Technical Errors in this Paper

This paper seems to have significant technical problems. Namely, Theorem 1 seems to be wrong. The calculus in the paper does not contain an equivalent of the cut rule. Thus, many formulae that are derivable under first-order semantics are not derivable using the presented calculus. Consider, for example, the DL KB conisting of these rules:

<math>\neg B(a)</math>

<math>A \sqsubseteq B</math>

These facts imply <math>\neg A(a)</math> under the standard first-order semantics. The calculus presented in the paper, however, does not derive this. In fact, the calculus does not contain any rules for dealing with negation.

[edit] Reply to Technical Errors in this Paper

First of all, sorry for my delay. I am one of the authors, and I did not realize such comments published here.

In this paper, an inferred TBox has been initialized, namely, <math>T^* = \{C \sqsubseteq D | \Sigma \models (C \sqsubseteq D)\} \cup \{P \sqsubseteq Q | \Sigma \models (P \sqsubseteq Q)\} </math> where <math>\Sigma</math> is a given DL KB. As a result, <math>A \sqsubseteq B</math> in the DL KB implies <math>\neg B \sqsubseteq \neg A</math>, which has been initialized in <math>T^*</math>, making <math>\neg A(a)</math> is derived by an inference rule of <math>\sqsubseteq_{T}</math>: If <math>\Gamma \vdash (C \sqsubseteq D)</math> and <math>\Gamma \vdash C(a)</math>, then <math>\Gamma \vdash D(a)</math> where <math>\Gamma</math> is the system w.r.t. the DL KB <math>\Sigma</math>. More details are shown in Section 4.2 of this paper.

As for certain formulae that are derivable under first-order semantics, I am afraid the KB, in which those formulae are involoved, itself is possibly unsatisfiable in epistemic interpretations, concerning semantical integrity constraints proposed in this paper. Theorem 1 shows up the soundness and completeness, on the assumption that KBs are satisfiable. As well-known, there exists unavoidable semantic discrepancy between DL and DB. So far, we should balance the trade-off, and this paper belongs to the direction of research which attempts to support ontology query answering on databases.

Welcome to more discussion. Thanks a lot.