wood@jfred.siemens.edu (Jim Wood) (07/03/90)
I'm interested in any information, positive or negative, about the LASER product from Bell Atlantic Knowledge Systems, particularly information concerning: - syntax of the rules and knowledge representation, - performance of applications, - product bugs and wierdnesses, and - maintenance and support from Bell people. Thanks very much for any help. Jim Wood P.S.: The version of LASER which I'm considering would run on SUN/UNIX platforms. Also, are any of the same kinds of comments available for NExpert Object and GoldWorks II? Thanks (again) --------------------------------------------------------------------------- Jim Wood [wood@cadillac.siemens.com] Siemens Research Center, 755 College Road East, Princeton, NJ 08540 (609) 734-6500
forrest@utkux1.utk.edu (Forrest M. Hoffman) (07/09/90)
>From: wood@jfred.siemens.edu (Jim Wood) >I'm interested in any information, positive or negative, >about the LASER product from Bell Atlantic Knowledge >Systems, particularly information concerning: > > - syntax of the rules and knowledge representation, > - performance of applications, > - product bugs and wierdnesses, and > - maintenance and support from Bell people. > >Thanks very much for any help. > > Jim Wood I have been using LASER for almost 2 years. At Oak Ridge National Lab, we are developing an expert system for analysis of solution-mineral interactions in geochemical/environmental problems. The knowledge representation capabilities of LASER are very good. User-defined relations for objects is a nice feature and the reverse links are provided, unlike in KES. The RPS rules use the same syntax as the KR module. We have not used NIE since our application requires only forward-chaining. There have been many bugs in LASER, but almost all of them have been cleared up. Whenever we have encountered a problem, we have contacted Bell Atlantic (Rafal or Ajay) and it has been fixed. They promptly sent us new versions. The maintenance and support are very good as far as I know. There are some drawbacks to LASER. First, the knowledge base is stored in an ASCII file so loading is slow; however, because it is ASCII, it is portable to any system and does not require compiling. This problem was overcome by faster machines and disk drives. Second, the rule execution can be slow since the inference engine works straight out of the rule objects (there is no compiling of rules). This could be a problem if your application is very large and you want extremely fast results; however, the fact that rules are not compiled means that you can design a rule-based system which can modify its own rules. This has great implications for learning or behavior adjustment. Third, everything is in memory so be sure your machine can handle it. Our knowledge base is getting very large and our system does so much that we are using almost all the 16MB in our DECStation 3100. Of course this depends on your application and the amount of information you have to keep around. We will just buy bigger hardware. We have combined our window interface and mathematical models with our expert system into one executable program. Publications of our work are currently in the review process. Forrest Hoffman forrest@esdhof.esd.ornl.gov ************************************************************************** * I speak only for myself........ * **************************************************************************