zbMATH — the first resource for mathematics

Reasoning and revision in hybrid representation systems. (English) Zbl 0702.68095
Lecture Notes in Computer Science, 422; Lecture Notes in Artificial Intelligence. Berlin etc.: Springer-Verlag. XII, 270 p. DM 42.00 (1990).
This book provides the reader with a broad and advanced understanding of knowledge-based systems, one of main research subjects of Artificial Intelligence. Particularly, the topics of reasoning and revision are emphasized. Based on the extensive investigation and analytical exploration of terminological formalisms and assertion formalisms along with other formalisms of representations, the extension has been made to a hybrid representation system (TF/AF). Then the reasoning aspect is focused in the light of subsumption, classification, and hybrid inference. Some contributing results have been achieved concerning the algorithms as well as their computational complexity. In addition, the problem of terminological cycles is handled with the recognition that terminological cycles can be useful to capture some meanings (e.g. intuitive semantics). That is impressive because it is usually ignored or avoided. Work done in this respect includes the evaluation of three semantic styles, the decidability of the subsumptions, and the extension of the inference algorithms. After the discussion of the representation formalism and inference, the very important issue of revision is then focalized. That is, the “changing world” demands changes of knowledge bases. An excellent survey of current approaches has been carried out, upon which it is shown that solutions for logical database updates and counterfactural reasoning can be reconstructed in the logic of theory change in terms of epitesmic relevance. Having developed a number of requirements for the revision, and realized the dissatisfaction of current solutions, the author adapts an approach aiming at minimal changes on the symbolic level, which possesses the closure property of revision operation. Besides, exploration is made to see how terminological revision can be performed by combining with the theory change approach in order to design an intuitive plausible and simple model of hybrid systems.
As a conclusion, the book is well organized, the subjects concerned are systematically described, and the corresponding treatments are rigorous, formal and insightful.
Reviewer: E.Kerre

68T30 Knowledge representation
68T35 Theory of languages and software systems (knowledge-based systems, expert systems, etc.) for artificial intelligence
68-02 Research exposition (monographs, survey articles) pertaining to computer science