SHRIF, a general-purpose system for heuristic retrieval of information and facts, applied to medical knowledge processing

摘要

We describe a completed project which, conceptually and technically, is a significant extension of IRUHS (Information Retrieval Using Heuristic Search) described in Findler (1979). SHRIF (System for Heuristic Retrieval of Information and Facts) is an information and fact retrieval system that emulates the natural strength and flexibility of humans in gathering new knowledge, grouping facts that belong together from some point of view, discovering and making use of partial similarities, handling uncertainties and incompletenesses, and making acceptable compromises when total consistency is not available. The system also exhibits user-friendliness in various ways. It receives new or modified data and queries in a high-level, English-like language. Users are not obligated to specify their information needs exactly and completely. They can express the different degrees of importance of the properties they expect the returned objects to have. They also declare their strength of belief that the named properties are indeed shared by the objects in question. The system can respond to less than complete or specific information requests. When the user is not quite satisfied with an answer, the input can be changed so that subsequent responses may be more suitable. A reasonably complex medical knowledge base was used in the development of the project.