On 26/04/2011 12:45 PM, John F. Sowa wrote:
> Ron,
>
> But the question is which tool is only appropriate to a "fraction"
> of the total problem.
I was think of ontology as a concept rather than statistical reasoning
or inference engines.
The actual software used is important at some level but the ability to
build successful reasoning processes that combine the rigidity of an
ontology with the flexibility that comes from absorbing conflicting
information from a variety of sources and sorting the relevant from the
information that only shares some of the vocabulary by accident, is
really the important bit.
> JFS
>>> [Watson] used ontology, but their major database was DB2 -- a good,
>>> old fashioned relational DB. For complex reasoning, they used Prolog.
>>> See the article (abstract below) about using Prolog to reason with
>>> information expressed in UIMA, which was another project by Ferrucci.
> RW
>> Perhaps this is an example of a typical enterprise's use of ontology.
>> A fraction of the total solution but key to driving the understanding of
>> the application domain's entire knowledge base.
> The original description logics were used in hybrid systems, in which
> the DLs served as the T-box (terminology component) and a richer version
> of logic served as the A-box (assertion component). But in the current
> version of the SemWeb, OWL is the only reasoning engine, and it was
> deliberately limited in its capabilities.
>
> Prolog is a much richer system, which can do everything that OWL can do
> plus a whole lot more. At our VivoMind company, we have a wide variety
> of reasoning tools, but we build them on top of Prolog. That's not
> possible with OWL.
>
> We also have a variety of other tools, some of which have been
> rewritten in C for better performance, but all of them were originally
> written in Prolog. We only translated them to C after we had tested
> them and were certain that they were critical for performance.
>
> Please remember that IBM was originally so enthusiastic about the SemWeb
> that they hired Guha -- the original designer of RDF. They also have
> many people on the Watson project who know OWL, such as Chris Welty.
> But when they wanted to build a complete semantic system, they did
> not choose RDF and OWL. They chose Prolog and UIMA. We use Prolog
> plus conceptual graphs.
>
> Also look at Experian -- one of the 3 largest companies that evaluate
> everybody's credit score. They do that in Prolog. It's so critical
> to their business that they bought Prologia, the company founded by
> the man who built the first version of Prolog.
>
> The simple fact is that you can use Prolog to implement the complete
> application, and Prolog has APIs that connect to mainstream software
> tools. That is why I say that Prolog is a step *forward* from OWL.
>
I look forward to the day when the application developer's guide for
Watson is on-line so we can see how to build new applications.
> John
>
> _________________________________________________________________
> Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
> Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
> Community Files: http://ontolog.cim3.net/file/work/OntologySummit2011/
> Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2011
> Community Portal: http://ontolog.cim3.net/wiki/
> (01)
_________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2011/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2011
Community Portal: http://ontolog.cim3.net/wiki/ (02)
|