- From: Harry Halpin <hhalpin@ibiblio.org>
- Date: Thu, 08 Dec 2005 14:25:20 +0000
- To: semantic-web@w3.org
While I appreciate Adrian's demonstrations, I'm just going to point out that while phrasing things in natural language explanations will clearly be useful for the semantic web, it doesn't correct - in fact it makes worse - the questions of ambiguity that I think were trying to be addressed by formal semantics to begin with. I don't think we have any choice but to assume any triple-store is not *just* the triple store but *the triple-store plus standardized inferences.* With vanilla XML is what you see is what you get, but in RDF that does not seem to hold all the time. Adrian Walker wrote: > > Danny -- > > You mentioned... > >> the issue of stores >> containing identical data responding differently to the same query >> (depending on their inference capability) > > > As you know, many of the inference systems currently in use take a > procedural approach -- e.g. forward chaining rule firing is > commonplace. Thus, the results from a query depend on which version > of which method is being used. > > Up till now, inference systems have mainly been used standalone, so it > has been just about possible to live with such a situation. > (Although, arguably, some major commercial inference-based projects > have failed because of such a procedural approach.) > > However, it's hard to see how this could continue to "sort of work" in > a (semantic) web context. The situation is going to get very messy > very fast, and no business person is going to trust results obtained > from such a software mess. E.g., a look at the jena-dev list almost > any day yields "unexpected inference" questions. > > Two fixes come to mind. > > Fix 1: Move all inferencing for the SW to *highly* declarative > reasoners that are compliant to a formal model theory that *defines* > what results must follow from *any* database-query pair, as in e.g. > [1] . Provide English explanations of results, at the non-programmer, > business level. > > Fix 2: Note that Fix 1 would require many vendors to trash their > products, and that is unrealistic in the short term. Persuade W3C and > other standards bodies to recommend an approach based on message > passing between inference systems -- in which each message *must* be > an English sentence that describes its meaning in the real world. > This approach is argued in more detail in [2]. > > How does that sound? Thanks in advance for comments. > > -- Adrian > > [1] Backchain Iteration: Towards a Practical Inference Method that is > Simple Enough to be Proved Terminating, Sound and Complete. Journal > of Automated Reasoning, 11:1-22. > > [2] Understandability and Semantic Interoperability of Diverse Rules > Systems > www.w3.org/2004/12/rules-ws/paper/19 > > > > INTERNET BUSINESS LOGIC (R) > Online at www.reengineeringllc.com > Shared, community use is FREE. > > Adrian Walker > Reengineering > PO Box 1412 > Bristol > CT 06011-1412 USA > > Phone: USA 860 583 9677 > Cell: USA 860 830 2085 > Fax: USA 860 314 1029 > > > >
Received on Thursday, 8 December 2005 14:25:42 UTC