An interpreter for large knowledge bases

• Authors: M. Nussbaum
• Venue: CSC '89 Proceedings of the 17th conference on ACM Annual Computer Science Conference
• Year: 1989

Achieving performance when both implicit and explicit information is present is an important issue in knowledge processing. As there has not yet been an adequate solution for the problem of large knowledge bases, most known solved problems deal with a limited amount of data. The alternative presented in this paper has proven effective in increasing the speed of knowledge processing for problems that manipulate large sets of facts and rules by orders of magnitude over conventional techniques.Most solutions have been focused on processing low granularity, i.e. one tuple at a time. Within the logic programming framework, an inference engine is proposed suited for large problems. It works with a high degree of granularity, i.e. a relation that permits a combination of a top-down and bottom-up evaluation strategy. It is top down in the way the Intentional Database (rules) is searched and bottom up in the sense the Extensional Database (facts) is processed. This technique, based on a reduction strategy, looks for a non-terminal node only on demand and saves partial results for future needs.The computational model was implemented. Benchmarks comparing the proposed interpretation scheme with other Prolog systems are presented and show a strong improvement in the performance.