User:Heron10/sandbox

Expert systems were among the first forms of artificial intelligence (AI) software. The systems attempted to capture the expertise of specialists to make the knowledge available to others. Initially, this methodology was not available to the general public because of mainframe computer costs and complex programming requirements. A variety of expert system building tools for personal computers were developed in the 1980's that made this methodology available outside of universities and large corporations. VP-Expert was one of these systems.

= Origins =

VP-Expert was first published by Paperback Software International in 1984 and upgraded in 1985 and 1987. In June 1991, ownership of VP-Expert was transferred to WordTech Systems, Inc. after Paperback Software was liquidated. In 1992 Wordtech supported an upgrade to version 3.0. VP-Expert was widely used in the 1980’s, but usage declined in the early 1990’s. The expert system shell was written in the C programming language. It was created and upgraded by Brian Sawyer, author of Programming Expert Systems in Pascal

= Features=

VP-Expert was a rule-based expert system in which the knowledge engineer entered if-then rules. A knowledge base was developed, a collection of rules pertaining to the expertise of an area of knowledge. Induction was more direct means of entering the expertise, using a set of examples and observations by the user. Currently, data mining is used to discover new knowledge in data unknown to the expert. Once the expert system was built, the expertise could be ported to end users on the IBM-PC. Current expert systems can be developed and ported to a range of machines from personal computers to mainframes. VP-Expert could receive input from spreadsheets and database software. Rules could be backward chaining or forward chaining. VP-Expert included hypertext, support for mathematical functions, and confidence factors for facts. The system also had a provision for fuzzy variables. With fuzzy variables the user was able to specify multiple answers to a question. In graphics mode, the developer could create screens that allowed the user to input data from images means of a mouse. Current expert systems provide the user with source code to build new features into the system.

= Applications =

VP-Expert has been utilized in various ways to fulfill needs in the medical field, law, business and applied science. It has been used to diagnose problems, to model and optimize systems, to provide training, and to automate repetitive tasks.

Medical uses were diagnosing gastrointestinal diseases, localizing the hippocampus in human brain from MRI scans automatically and training nurses. A Sentencing Advisor was used by the legal community. Two scientific uses were a system to balance airport capacity and noise mitigation, and nuclear reactor safety analysis. A business application supported analysts in the area of Mergers and Acquisitions.

= Context =

Aspects of other early expert system pc shells that were widely used, Level 5 Object, and Exsys Professional, may be compared to VP-Expert during the period from 1989-1992. LEVEL-5 Object had the most flexible inference methods, utilizing forward chaining, backward chaining, blackboard, and multiple hypotheses. EXSYS had a similar level of flexibility. VP-Expert primarily used backward chaining and forward chaining rules. Level-5 Object ran on Apple Macintosh, DEC VAX and IBM Mainframe in addition to the IBM PC, while VP-Expert ran only on the IBM PC VP-Expert had good demonstration capabilities, displaying the contents of the knowledge base knowledge base that it was currently searching, the active inference process, and the current state of working memory. EXSYS showed only the rules that were currently being applied. LEVEL-5 Object had minimal demonstration capabilities. These systems needed to communicate well with the end-users to answer their questions and to give advice. Both LEVEL-5 Object and VP-Expert had good explanation capabilities while EXSYS was limited in this area. All three packages could be integrated with Dbase, Lotus 1-2-3, and ACSII files. LEVEL-5 Object could also be integrated with FOCUS. All these systems used confidence calculations for the rules. The above expert system shells are no longer supported. CLIPS and JESS are expert systems currently in wide usage. CLIPS is public domain software, with source code freely available, and running on any machine that has an ANSI compliant C compiler. JESS is a clone of CLIPS, written entirely in JAVA. = References =