The Vex-93 Environment as a Hybrid Tool for Developing Knowledge Systems with Different Problem Solving

The paper describes VEX-93 as a hybrid environment for developing knowledge-based and problem solver systems. It integrates methods and techniques from artiicial intelligence, image and signal processing and data analysis, which can be mixed. Two hierarchical levels of reasoning contains an intelligent toolbox with one upper strategic inference engine and four lower ones containing speciic reasoning models: truth-functional (rule-based), probabilistic (causal networks), fuzzy (rule-based) and case-based (frames). There are image/signal processing-analysis capabilities in the form of programming languages with more than one hundred primitive functions. User-made programs are embeddable within knowledge basis, allowing the combination of perception and reasoning. The data analyzer toolbox contains a collection of numerical classiication, pattern recognition and ordination methods, with neural network tools and a data base query language at inference engines's disposal. VEX-93 is an open system able to communicate with external computer programs relevant to a particular application. Metaknowledge can be used for elaborate conclusions, and man-machine interaction includes, besides windows and graphical interfaces, acceptance of voice commands and production of speech output. The system was conceived for real-world applications in general domains, but an example of a concrete medical diagnostic support system at present under completion as a cuban-spanish project is mentioned. Present version of VEX-93 is a huge system composed by about one and half millions of lines of C code and runs in microcomputers under Windows 3.1. 1 Project partially supported by a grant (1993) from the Alexander von Humboldt Foundation (Bonn, Germany).