| Abstract | Four principal Web rule issues constitute our starting points: I1) Formal knowledge representation can act as content in its own right and/or as metadata for content. I2) Knowledge on the open Web is typically inconsistent but closed 'intranet' reasoning can exploit local consistency. I3) Scalability of reasoning calls for representation layering on top of quite inexpressive languages. I4) Rule representation should stay compatible with relevant Web standards. To address these, four corresponding essentials of Web rules have emerged: E1) Combine formal and informal knowledge in a Rule Wiki, where the formal parts can be taken as code (or as metadata) for the informal parts, and the informal parts as documentation (or as content) for the formal parts. This can be supported by tools for Controlled Natural Language: mapping a subset of, e.g., English into rules and back. E2) Represent distributed knowledge via a module construct, supporting local consistency, permitting scoped negation as failure, and reducing the search space of scoped queries. Modules are embedded into an 'Entails' element: prove whether a query is entailed by a module. E3) Develop a dual layering of assertional and terminological knowledge as well as their blends. To permit the specification of terminologies independent of assertions, the CARIN principle is adopted: a terminological predicate is not permitted in the head of a rule. E4) Differentiate the Web notion of URIs as in URLs, for access, vs. URNs, for naming. A URI can then be used: URL-like, for module import, where it is an error if dereferencing does not yield a valid knowledge document; URN-like, as an identifier, where dereferencing is not intended; or, as a name whose dereferencing can access its (partial) definition. |
|---|
