- From: John F. Sowa <sowa@bestweb.net>
- Date: Fri, 27 Jun 2008 14:46:47 -0400
- To: "[ontolog-forum]" <ontolog-forum@ontolog.cim3.net>
- CC: semanticweb@yahoogroups.com, welty@us.ibm.com, public-semweb-lifesci hcls <public-semweb-lifesci@w3.org>, semantic_web@googlegroups.com
I just wanted to add some comments that I forgot to make in my previous note in this thread. Following is what I did say: There are also many cases where a metalevel reasoner can make orders of magnitude *improvement* in performance at the object level. Among them are various kinds of compilers that reorganize or specialize the axioms at the object level before proceeding with the proof. Two important points to add are the following: 1. Most commercially successful methods for handling defaults, exceptions, negation-as-failure, belief revision, and related methods of nonmonotonic reasoning already depend on a metalevel reasoner. 2. Unfortunately, that reasoner is located in the head of some programmer, knowledge engineer, or database administrator who is forced to anticipate all possible options and write special-purpose code or axioms to handle them. I also cited the work by Andersen et al. as a valuable metalevel approach for compiling from a very general, very expressive version of logic (CycL) to a more specialized logic used in conjunction with a relational database: Peterson, Brian J., William A. Andersen, & Joshua Engel (1998) "Knowledge bus: generating application-focused databases from large ontologies," Proc. 5th KRDB Workshop, Seattle, WA. http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-10/ The authors used Cyc as a powerful tool for knowledge acquisition and testing. After they had used Cyc to develop their ontology, they used their compiler to translate the Cyc axioms to simpler statements (Horn-clause logic and database constraints). The result ran much faster than the original Cyc version, and it did not require the Cyc system at run time. Unfortunately, Doug Lenat, the director of the Cyc project, lost a fantastic opportunity. Instead of promoting Cyc as a powerful development tool for databases and knowledge bases that could run independently of Cyc, Doug took a very negative, defensive position. He wanted everybody to use Cyc, and he did not want people to compile Cyc axioms to systems that did not run the Cyc software. But by insisting on that narrow use of Cyc, Lenat lost a chance to move Cyc into the mainstream of commercial development and eventually the Semantic Web. I believe that commercial success of semantic technology depends critically on moving metalevel compilers out of the heads of professional knowledge engineers. We cannot expect large numbers of people to learn and use RDF(S) + OWL + RuleML + SPARQL or any similar notations. But we can expect domain experts to work with automated and semiautomated knowledge compilers that can generate the appropriate run-time code. John Sowa
Received on Friday, 27 June 2008 18:47:30 UTC