Official Project-Website
|
|
(IST project IST-FET-2001-37004) Official Project-Website |
Description logics are well-known tools for handling ontological knowledge and play an important role in building the Semantic Web. For the Ontology Layer of the Semantic Web, the Web Ontology Language (OWL) is a W3C recommended standard, with sublanguages OWL Lite, OWL DL, and OWL Full; the former two are based on the description logics SHIF(D) and SHOIN(D), respectively. Current and future efforts in building the Semantic Web are aimed at the Rules, Logic, and Proof Layers on top of the Ontology Layer. As they should offer sophisticated representation and reasoning capabilities, this requests, in particular, the need to integrate the Rules and the Ontology Layer.
Several proposals for combining description logics with rule-based languages exist. One are description logic programs (dl-programs, for short), representing a novel method to couple description logics with nonmonotonic logic programs. Semantically, they fully support encapsulation and privacy of the components, in the sense that logic programming and description logic reasoning are technically separated and only interfacing details need to be known.
Roughly speaking, a dl-program consists of a description logic knowledge base L (in SHIF(D) or SHOIN(D); however, other description logics could be used as well) and a finite set P of generalized logic-program rules, called dl-rules. Such rules are similar to usual rules in logic programs with negation as failure, but may also contain queries to L in their bodies. Importantly, such queries may involve an input from P to L, and hence a bidirectional flow of information between P and L is facilitated. Thus, dl-programs allow for building rules on top of ontologies, but also, to some extent, building ontologies on top of rules.
Two basic types of semantics have been defined for dl-programs: in [5], a generalization of the answer-set semantics for ordinary logic programs is given, and in [4], a generalization of the well-founded semantics. In fact, two versions of the answer-set semantics for dl-programs are introduced, namely the weak answer-set semantics and the strong answer-set semantics. Both semantics coincide with usual answer sets in the case of ordinary normal programs. Every strong answer set is also a weak answer set, but not vice versa. The two notions differ in the way they deal with nonmonotonic dl-queries. While the answer set semantics resolves conflicts by virtue of permitting multiple intended models as alternative scenarios, the well-founded semantics remains agnostic in the presence of conflicting information, assigning the truth value false to a maximal set of atoms that cannot become true during the evaluation of a given program.
In general, our approach enhances the restricted reasoning capabilities of description logics by the features of a declarative, nomonotonic rule language, but keeps decidability at the same time. Description logic knowledge bases lack the possibility of specifying the closed-world assumption, which is acknowledged as an important reasoning principle for inferring negative information. Weakly negated dl-atoms, querying a concept or role, together with the option of feeding back the result to the respective complement, enables to explicitly close parts of the DL-KB. In a similar fashion, models can be minimized or maximized, which furthermore may be exploited to support default reasoning on top of a DL knowledge base.
Several small examples on the prototype website illustrate the capabilities and benefits of dl-programs. The "Artist"-example shows how to minimize the extension of a concept. According to the corresponding ontology, Artists are either Singers or Painters. Under a suitable closed world assumption, the question who is a singer or a painter should be answered. To this end, the minimal class extensions of the ontology on the individuals in the ontology are considered. A more complex scenario is the "Web Services"-example, which searches for web services suggesting White Wine for a specific meal. This dl-program uses an adapted version of the publicly available OWLS-TC Web Services collection.
These two and further examples are available (and ready to be evaluated) at the prototype website.
NLP-DL Prototype, including some example dl-programs along with the possibility to specify custom programs.
Literature:
page mainted by Roman Schindlauer;