Synthesizing constraint expressions
Communications of the ACM
The inference of domain structure from informal process descriptions
ACM SIGART Bulletin
IJCAI'79 Proceedings of the 6th international joint conference on Artificial intelligence - Volume 2
An overview of recent data base research
ACM SIGMIS Database
| Hi-index | 0.02 |
This report presents a novel network-like representation for information, called "constraint expressions" (CE). CE makes use of some of the knowledge-representation techniques developed by Artificial Intelligence research. A CE network consists of points (which represent classes of objects) interconnected by constraints (which represent the relationships which are known to hold among the classes). All constraints are defined in terms of six primitive ones. The data in a CE network is accessed by propagating various kinds of labels through it: Each constraint can be viewed as an active process which looks for certain patterns of labels on some of its attached points, and then propagates new labels to other points when such patterns occur. The CE representation provides several significant features which are not found in most current data models. First, the same mechanism is used to represent "general" as well as "specific" information. For example, "The sex of Jane Smith is female" is specific, while "Every person has a unique sex which is either ''male'' or ''female'' " is general. Second, CE''s label-propagation procedure implements logical consistency checking: Data base integrity can be maintained by checking all new data for consistency with the existing information. Since the data base can contain general information (representing a "semantic model" of the data base''s application domain), new specific data can be rejected if it is inconsistent with either other specific data or with the general information. Also, the general information can itself be checked for internal consistency. Third, the CE representation is sufficiently modular and well-defined so that it has precise formal semantics, which insures that CE''s definition contains no hidden ambiguities or contradictions. Fourth, CE''s modularity allows the label propagations to be done in parallel, so that parallel hardware can be used to full advantage.