A Sorted-Graph Unification Approach to the Semantic Web

Summary form only given. Today, the main challenge faced by the Semantic Web (SW) project is to process knowledge expressed in terms of sorted graphs, not only for general-purpose WWW and service-oriented computation, but also and most importantly, for inference and learning. To wit, the mainstream formalisms used in SW e.g., Description Logic (DL), Resource Description Framework (RDF) are essentially tools for describing sorted graphs. Yet, sorted graphs have been used for decades now in Artificial Intelligence (Al), Knowledge Representation (KR), and Natural Language processing (NLP), in the form of attribute-inheritance concept networks. Among such approaches, one has consisted in extending unification of logical terms as used in Automated Theorem-Proving (ATP) and [Constraint] Logic Programming ([C]LP) to work on graphs representing conceptual information. This is the case of Order-Sorted Feature (OSF) unification technology, which is based on viewing all data as a constraint formula representing a sorted graph. Thus, inference based on such data amounts to constraint solving in a formal logic of constraints over order-sorted triples as the basic building blocks which incidentally have been used in the Warren Abstract Machine (WAM) technology since the 70's in Prolog compilers. In this talk, I will review the essence of the OSF formalism in relation to mainstream SW formalisms for the purpose of making the SW a reality. Doing so, I will show examples using LIFE, a CLP language combining logical (Horn) rules and functional (rewrite) rules over OSF graph constraints. LIFE stands for "Logic, Inheritance, Functions, and Equations". LIFE'S logical and functional rules boost the expressive power of OSF graphs by allowing them to be constrained with user-definable predicates and functions.