% talk abstract for invited talk at Stanford U. Semantic Web Seminar Mar. 2006 % hosted by Prof. Michael Genesereth and his student Mike Kassoff Time and Location: 12:15pm on Wed., March 29, 2006; in Gates 2A (near 220) The Production Logic Programs Approach: KR Foundations for Semantic Rules on the Web Benjamin Grosof Assistant Professor of Information Technology, Massachusetts Institute of Technology, Sloan School of Management Co-Founder, RuleML Initiative Co-Editor, Semantic Web Services Initiative http://ebusiness.mit.edu/bgrosof Abstract: The use of knowledge-based techniques for e-services over the web has recently grown rapidly as a commercial area, focused largely on unstructured forms of knowledge, notably text. A new wave of knowledge-based technology focused on interoperable structured forms of knowledge, notably rules in combination with databases and ontologies, is exploding as a research and standardization area. We overview Production Logic Programs (PLP), our latest enhancement of the foundational knowledge representation (KR) approach at the heart of this new wave. This presentation will focus on the key concepts, theory, and techniques and, briefly, the fast-maturing standardization scene, emerging business applications, and business value analysis for rule-based semantic services -- much of which we have pioneered. W3C has recently launched a full blown standards effort in this area: Rule Interchange Format. PLP represents web rules that are semantically interoperable between all four of the currently most commercially important families of rule systems: production rules, relational database management systems, event-condition-action rules, and Prolog. The PLP approach combines high degrees of expressiveness, scaleability, and incremental/modular implementability. The PLP KR is based on declarative logic programs, extending Datalog with several major web-izing and expressive features. These features include, notably, a new approach to actions and tests via procedural attachments -- in the manner of production rules but with declarative semantics (which production rules previously lacked). Other features include default negation and Courteous prioritized conflict handling, which enable robust knowledge merging. The PLP approach includes smoothly and powerfully combining rules with ontologies of several commercially important kinds, including OWL, first order logic (FOL), object-oriented (OO), database schemas, XML schemas, and RDF schemas. The PLP approach has been piloted and evolved in a number of important rule formats, tools, and efforts -- including the influential RuleML standards design, its special case SWRL, and the closely related Semantic Web Services Framework; the SweetRules tool platform which supports those; the related WSML/WRL and IBM CommonRules; and most lately W3C RIF. An advanced set of PLP capabilities have already been implemented in open source and used to prototype a variety of e-business application scenarios. We discuss a number of directions for future research. Bio: at http://ebusiness.mit.edu/bgrosof/#Bio Additional Details (Abstract, continued): Rule-based semantic services techniques promise much deeper and cheaper business process automation and communication, enabling much expanded e-services, both intra- and inter- enterprise. Semantic rules are especially good for representing and automating many kinds of policy-flavor knowledge and a variety of services tasks. Their high level of conceptual abstraction facilitates specification and understanding by non-programmers. Generalizing database views and queries, semantic rules are the most expressively powerful form of structured knowledge that is practical for large scale deployment on the web, and offer much greater knowledge reuse. They enable loose-coupled integration of heterogeneous rule-based applications/services -- both forward-chaining/event-driven and backward-chaining/query-driven -- and distributed creation of large knowledge bases. ("Semantic" means having declarative interoperability in the sense of knowledge representation and database theory, independent of implementation/control-algorithm.) Our foundational work in semantic rules and their use for semantic services includes: 1. the dominant approach to standardizing interoperable rules in W3C and OMG, based on XML declarative logic programs (LP) cf. RuleML; 2. core knowledge representation theory and techniques, notably extensions of LP for procedural attachments for actions and queries, prioritized conflict handling and robust merging, and incorporation of ontologies -- embodied in the leading open source platform/toolkit (SweetRules); 3. the influential Semantic Web Services Framework design approach; and 4. application-specific techniques and prototypes in e-commerce (e.g., contracts, advertising, procurement, problem handling), information integration/mediation (e.g., financial and travel), and authorization/trust (e.g., security and confidentiality). Work supported in part by a DARPA Agent Markup Language program award and the Center for eBusiness @ MIT Vision Fund.