RuleML for Object-Relational Knowledge Representation on the Web
Harold Boley
AI has used relational and object-oriented knowledge representations, which have been complemented by representations of uncertainty. The (Semantic) Web is establishing an infrastructure for building large-scale, distributed knowledge bases. Languages for knowledge representation on the Web can integrate the relational ('rule') paradigm and the object-oriented ('frame') paradigm. Relational languages are layered from Datalog and Horn logic up to first-order, higher-order, and modal logics. Object-oriented languages include the Resource Description Framework (RDF), JSON, Notation3 (N3), and Frame logic (F-logic). In F-logic and W3C's Rule Interchange Format (RIF), frames are defined entirely separately from relation applications. In RuleML, these fundamental notions are integrated by permitting applications with optional object identifiers (OIDs) and, orthogonally, arguments that are positional or slotted. The resulting positional-slotted, object-applicative (psoa) terms have recently been endowed with a new formal semantics, reducing the number of RIF terms. Rule-defined frames are combinable with the ontology paradigm as follows: First, OID types (sorts, classes) can be organized into a subClassOf taxonomy (a 'light-weight' ontology), e.g. employing RDF Schema. Second, object descriptions on the schema/signature level correspond to description logics, the foundation of the new Web Ontologies Language (OWL 2). RuleML has used the first method to permit order-sorted typing of objects and rule variables, and is researching the second method based on RIF's OWL compatibility approach. RuleML has also explored representations of uncertain knowledge, e.g. in the Fuzzy RuleML Technical Group and with the RIF Uncertainty Rule Dialect (RIF-URD).