Bertaud Gounot Valérie, Donfack Valéry, Lasbleiz Jérémy, Bourde Annabel, Duvauferrier Régis
Unité Inserm U936, IFR 140IFR 140, Faculté de Médecine, University of Rennes 1, France.
Stud Health Technol Inform. 2011;169:714-8.
Expert systems of the 1980s have failed on the difficulties of maintaining large rule bases. The current work proposes a method to achieve and maintain rule bases grounded on ontologies (like NCIT). The process described here for an expert system on plasma cell disorder encompasses extraction of a sub-ontology and automatic and comprehensive generation of production rules. The creation of rules is not based directly on classes, but on individuals (instances). Instances can be considered as prototypes of diseases formally defined by "destrictions" in the ontology. Thus, it is possible to use this process to make diagnoses of diseases. The perspectives of this work are considered: the process described with an ontology formalized in OWL1 can be extended by using an ontology in OWL2 and allow reasoning about numerical data in addition to symbolic data.
20世纪80年代的专家系统在维护大型规则库的难题上遭遇了失败。当前的工作提出了一种基于本体（如NCIT）来实现和维护规则库的方法。这里所描述的针对浆细胞疾病专家系统的过程包括子本体的提取以及生产规则的自动全面生成。规则的创建并非直接基于类，而是基于个体（实例）。实例可被视为通过本体中的“约束”正式定义的疾病原型。因此，利用这个过程进行疾病诊断成为可能。本文还探讨了这项工作的前景：用OWL1形式化的本体所描述的过程可以通过使用OWL2中的本体进行扩展，并且除了符号数据之外还能对数值数据进行推理。
