Reference: J. Rice, A. F. OKBC, A Rich API on the Cheap. Knowledge Systems Laboratory, Palo Alto, CA, April, 1998.
Abstract: The development of sophisticated knowledge bases and knowledge-based applications has traditionally been hampered by the idiosyncratic implementation of both the knowledge in these systems and the representation systems in which they are implemented. The lack of standards and the desire of each implementor to exploit different technologies has made it very difficult to leverage existing knowledge bases or existing tools, since these tools have traditionally been very tightly bound to a single representation system. Because representation systems differ in significant and interesting ways, it seems unlikely that we will ever be able to have a grand unified representation system that will ensure portability. Nevertheless we would like to avoid the terrible costs associated with reinventing tools and representation systems, just as we would like to be able to reuse existing knowledge bases. The desire for reuse becomes especially strong as our knowledge bases grow and we would like to extract the data-rich, though often ontologically impoverished, OKBC (Open Knowledge Base Connectivity) is an API (Applications Programming Interface) (Chaudhri et al. 1998) specifically designed to allow developers to access knowledge content in a wide variety of systems, and to allow application developers to develop sophisticated, portable and reusable applications. However, in order for OKBC to be successful, it must rely on bindings for OKBC being provided for a large number of content providers as well as relying on the API being simple, powerful and convenient enough for application developers to see a pay-off in using it. The desire to provide a powerful and flexible API is clearly incompatible with the desire to make it easy to provide bindings for systems. In this paper we show how OKBC has been designed specifically to overcome this
Full paper available as pdf.