Computer aids for chemists

Abstract

The goal of this ongoing project is to build a family of expert systems that can be used easily for education and research in different disciplines to solve problems in both experimental work and development of theories. Important characteristics of the systems are described. There are three classes of expert systems: conventional, automatic deductive (with symbolic and non-symbolic subclasses), and adaptive learning (with conventional and automatic subclasses). Automatic deductive and automatic adaptive learning systems are not based on the state-of-the-art and can predict unprecedented events. A basis for many problems is the construction, manipulation, and evaluation of graphs. There are two automatic deductive and two automatic adaptive learning systems capable of solving four classes problems common to many disciplines. Two interfaces called INTERACTIVE USER and SYSTEM together provide user-friendly interfaces for different subjects and varying levels of expertise. Only the INTERACTIVE USER interface need be changed for new applications. Nine service expert systems perform operations such as literature searches, high-speed information management, curve-fitting and data description. Operational prototypes exist for ten of the thirteen systems. CURFIT (curve-fitting) and FRANS (function recognition and numerical solutions) systems are available for production use on IBM-PC/AT machines and their clones. The use of the CURFIT and FRANS systems is illustrated with some simple examples.