Knowledge Processing under Information Fidelity

Abstract

XSPIRIT is a professional expert system-shell for knowledge acquisition, inference and response using conditional logic and probability. Composed conditionals on propositional variables with finite domain are the communication tool between the user and the knowledge base, making the process of acquisition, inference and query comfortable and intelligible. XSPIRIT allows partial rather than complete information about the knowledge domain and supplements missing parts by the principle of information fidelity. By virtue of evident temporary information, knowledge undergoes a well-defined adaptation process, respecting this principle again. The construction and transformation of probability distributions as developed here, allow acquired knowledge, remaining uncertainty and strength of inference to be measured in the information units [bit]. XSPIRIT allows large-scale applications with hundreds of composed conditionals and umpteen variables.