知识获取的多层证据网络模型

本文从诊断问题求解知识的一种组织形式——证据网络出发,讨论了将领域知识转换成证据网络的各种策略,并由此提出了知识获取的多层证据网络模型。根据该模型我们开发了交互式知识获取系统KAS-CEI,并将该系统用于汽车发动机故障诊断专家系统(AEFDES)知识库的建造。

A Multi-Level Belief Network Model for Knowledge Acquisition

A Knowledge Acquisition Model for diagnostic problem solving, termed multilevel belief network (MBN), is presented. It classifies the diagnosis knowledge into difficult and easy ones described by causal belief network and diagnostic belief network respectively. In MBN, three "knowledge chunk" strategies are employed, that is, goal first strategy, evidence first strategy and case synthesis strategy. The goal first strategy links the belief network first from a goal of diagnostic problem solving, the evidence first strategy links the belief network first,from facts about diagnostic problem solving and case synthesis strategy forms the belief network through an analysis of a series of diagnosis examples. The advantage of these strategies lies in keeping the coherence and would be easily accepted by domain experts during knowledge acquistion. In order to illustrate the application of MBN, an interactive knowledge acquisition system KAS-CEI based on MBN is given. KAS-CEI has several independent modules such as knowledge transferring module, concept analysis module, completeness checking module, knowledge base generation module and knowledge base testing module. It supports various inference mechanisms including deductive inference, abductive inference and mixture inference, and it is also capable of forming various knowledge bases with the rule, frame and semantic network representation.