Knowledge representation in PARKA

Abstract : This paper describes a frame-based knowledge representation system called PARKA which runs on the Connection Machine. The PARKA language was designed to take advantage of the Connection Machine architecture in order to provide extremely fast query processing capabilities. While the primary goal in PARKA's design was to explore the use of massive parallelism in symbolic knowledge representation, we have also tried to respond to many of the criticisms and to take advantage of many of the advances documented in the literature of frame languages. PARKA provides a principled approach to multiple inheritance based on the work of Touretzky (21), and a new mechanism for the representation of part/whole relations. In addition, we have made efforts to make the intended meanings of PARKA representations clear from the perspective work in the philosophy of language. The result is practical representation system that can run very quickly on parallel hardware. A specification of the syntax of the PARKA language is provided as an appendix.