- From: Francis McCabe <fgm@fla.fujitsu.com>
- Date: Tue, 28 May 2002 22:17:58 -0700
- To: "Champion, Mike" <Mike.Champion@SoftwareAG-USA.com>
- Cc: www-ws-arch@w3.org, "David W. Levine" <dwl@watson.ibm.com>
Mike: Glad you liked the post. I'll try to show light on some of the issues you raised. > > Thanks for a very informative post! > > One question: you mention that the semantic web builds on predicate > logic > .... > >> predicate logic makes an admirable graph notation and >> Ontology is the discipline that >> has had most to say about these semantic nodes. > > but does not try to assert a global set of axioms ... > >> One point to note about the Semantic WEB approach that is very >> interesting: viewed as an Ontology building system (which it is) >> the vision is that there will not be a single definitive Ontology, >> or even a single `upper ontology'. Instead, there are likely to >> be many ontologies and many cross links between them: its an >> ontology web rather than an ontology tree. > > My (limited, amateurish) understanding of the role of formal logic in > human > affairs is that the big challenge is to come up with a consistent set of > axioms on which to perform valid inferences but is rich enough to be of > practical significance. As a practical matter, this has only happened > in a > limited number of disciplines, and required centuries of effort. The > lack > of consistency renders the exercise utterly futile; for example, the > oft-cited anecdote (or joke, or whatever it really is): The main reason that centuries of effort were required is that they didn't have computers ;-) > > Bertrand Russell mentioned that a false proposition implies any > conclusion. A student raised his hand and said "given that 1 = 0, > prove that you are the Pope". Russell immediately replied, > "Add 1 to both sides of the equation: then we have 2 = 1. > The set containing just me and the Pope has 2 members. > But 2 = 1, so it has only 1 member; therefore, I am the Pope." There isn't anything fundamentally different from writing an axiom or writing a program. You can also create havoc with Java ;-) > > So, what hope is there that useful logical inferences could be made > using > cross-linked ontologies that have not undergone rigorous scrutiny to > ensure > their mutual consistency? Or is the type of inference required to reach > useful conclusions on the semantic web of a more heuristic variety than > textbook logic? For example, if one finds vastly more RDF pathways that > lead to the conclusion that Karol Wotilya is the Pope than one finds > that > conclude the Bertrand Russell is the Pope, is that good enough? (FWIW, a > Google search to dig up the Bertrand Russell joke got a lot more hits > than a > Google query to find out the current pope's given name!). A `normal' WEB bot looks for strings that it is interested in. The big problem with that is that it will register as a positive hit any web page that has the text "this page is not intended for adults" as a suitable place to go for adult entertainment. A parallel search using concept maps; especially where the concept map is written in a reasonable powerful logic, will be able to take into account the `little words' like `not' and `some' etc. However, you're still not guaranteed a perfect match. On the other hand, if the match IS good then you're in much better shape than a normal WEB bot as you've also got all those pointers to web services. > > >> The WS-Arch WG could, and should, make its requirements clear to the >> SWI. It is hard to imagine the W3C having two competing >> technologies for describing semantics. > > I personally think that given the state of immaturity of semantic > theory and > technology, and given the limitations of RDF and DAML that you > mentioned, > that it WOULD be quite appropriate for the W3C to have competing > technologies for describing semantics. This seems like a time for > exploration, and standardization can come when we understand what really > works well. The W3C works best as a forum for coordinated exploration > of > promising alternatives in a way that minimizes gratuitous > interoperability > differences. It does not work well when it tries to do "computer > science by > committee" that immediately leads to standardization. > Well, I have an axe to grind here. My colleagues and I have been working on strongly typed logics for a number of years. But, I still don't think that the W3C will have more than one semantic notation. I do think that its entirely appropriate for web services to send a strong message to the semantic web folk -- call it a wake up call!
Received on Wednesday, 29 May 2002 03:09:13 UTC