rich@artada.lerc.nasa.gov (Richard S.) (08/10/90)
As everyone is well aware, one of the biggest problems associated with any computer program, including expert systems, is how should they be documented. I am currently working on one method of documentation which includes using a hypertext software package (GUIDE 3, by Owl International), for my current expert system. This is being completed as a special topic in regards to my Graduate Work at the University of Akron. I have a few ideas of the kind of information, should be included. I would like to hear from others, who have developed expert systems, in regards to what kind of documentation efforts were considered, if any. Also, if any one has ideas of information that should be included/excluded. Also interested in any examples of good/bad documentation efforts. Any references to books, articles or papers associated with this topic??? Any and all comments are welcome. Please send e-mail to R3RFS@VM1.CC.UAKRON.EDU or reply here, I am sure that others are interested in this also. -- Rich Schlegelmilch | r3rfs@vm1.cc.uakron.edu Electrical Engineering Dept. | rich@artada.lerc.nasa.gov The University of Akron | Akron, OH 44325 | (216) 972-7649
hin@tnoibbc.UUCP (Hin Oey) (08/14/90)
In article <1990Aug10.124331.9222@eagle.lerc.nasa.gov> r3rfs@vm1.cc.uakron.edu (Richard Schlegelmilch) writes: >As everyone is well aware, one of the biggest problems associated with >any computer program, including expert systems, is how should they be >documented. > >I am currently working on one method of documentation which includes using >a hypertext software package (GUIDE 3, by Owl International), for my current >expert system. This is being completed as a special topic in regards >to my Graduate Work at the University of Akron. It is interesting to hear somebode talking about hypertext in relation to expert systems. At present I am finishing an expert systems, which uses a seperate hypertext package as an why/explain. In relation to maintaining software I have had a nice discussion in (mis)using SGML for preparing and maintaining source code. Kind Regards, Hin Oey -- Institute TNO for Building Materials and Structures (TNO-IBBC) Hin Oey Expert Systems Group PO BOX 49 tel: +31 15 842018 UUCP : ..!hp4nl!tnoibbc!hin 2600 AA Delft fax: +31 15 843990 USENET: hin@tnoibbc The Netherlands -- Institute TNO for Building Materials and Structures (TNO-IBBC) Hin Oey Expert Systems Group PO BOX 49 tel: +31 15 842018 UUCP : ..!hp4nl!tnoibbc!hin 2600 AA Delft fax: +31 15 843990 USENET: hin@tnoibbc The Netherlands
dahl@bcstec.UUCP (Mark Dahl) (08/17/90)
In article <1990Aug10.124331.9222@eagle.lerc.nasa.gov>, rich@artada.lerc.nasa.gov (Richard S.) writes: > As everyone is well aware, one of the biggest problems associated with > any computer program, including expert systems, is how should they be > documented. > > I am currently working on one method of documentation which includes using > a hypertext software package (GUIDE 3, by Owl International), for my current > expert system. > > I have a few ideas of the kind of information, should be included. > I would like to hear from others, who have developed expert systems... I'm looking into using Hypertext as part of an overall maintenance environment for a large set of expert systems. The hope is that hypertext can informally represent maintenance-related knowledge useful to human readers. This information can be layered on top of existing rule-based systems without affecting validated "source code". If hypertext nodes & links can be typed (or at least labelled), the "hyperbase" can later be formally merged with the rules to create a frame-base or semantic net, providing a platform for smarter knowledge acquisition, testing, and maintenance tools. Unfortunately, many commercial hypertext tools only provide untyped links and nodes that only form vague associations. I seen a little of GUIDE. It seems limited in expressiveness but maybe you can make up the difference with its programming language (?) As for as the kinds of information to include, our focus is on traceability to original knowledge sources. The other major area is documenting implicit knowledge useful during maintenance. Of course, what to include depends on how rich your knowledge representation is to begin with. For instance, many rule-based expert systems only provide attributes and values. The entities, relationships, classes, prototypes, etc. are nowhere to be found! The goal should be to generate documentation from expert systems. Other issues worth considering are versioning and support for collaborative work. A few references that might be useful in evaluating hypertext products are: "Shadow, fusing Hypertext with AI", by Patricia Carando, in IEEE Expert, Winter 1989. "Reflections on Notecards: Seven Issues for the Next Generation of Hypermedia Systems", by Frank G. Halasz, in CACM, July 1988. "Hypertext: An Introduction and Survey", by Jeff Conklin, in IEEE Computer, September 1987. --- Mark Dahl Boeing Computer Services Renton, Washington Scientific Data Systems - AI Specialist e-mail: dahl@bcstec.boeing.com -- Mark Dahl Boeing Computer Services Renton, Washington Scientific Data Systems - AI Specialist dahl@bcstec.boeing.com ...!uunet!bcstec!dahl