On Fri, October 12, 2012 11:49, John F Sowa wrote:
> ...
> OWL was optimized for one specialized part of one
> very specialized kind of reasoning: the time required to classify
> some item in a hierarchy. By imposing tight restrictions on the
> expressive power of a DL language, OWL can guarantee that operation
> is (a) decidable and (b) computable in polynomial time. (01)
> For some purposes, that property is useful. But in the grand scheme
> of Tim's vision, that issue is so tiny that he never mentioned it in
> *any* DAML report. ... (02)
> For anybody who is concerned about complexity and execution time in
> a large ontology, please look at the Cyc reports. It has a hierarchy
> of over 600,000 concepts. And the Cyc people (which include Doug F)
> have said that decidability was *never* a major concern. They have
> been working on very large reasoning problems for the past 28 years,
> and they have *never* encountered a problem for which the kinds of
> restrictions imposed by OWL were useful or desirable. (03)
John, if you are referring to restrictions on expressibility, i agree; but
if you are referring to restrictions on reasoning, i beg to differ. (04)
You just said that OWL was optimized for a specialized kind of reasoning.
Cyc knows specialized kinds of reasoning very well. It does not have a
single monolithic inference engine, but hundreds of inference modules
that are each optimized for a specialized kind of reasoning. (05)
OWL guarantees that the time required to classify some item in a hierarchy
is computable in polynomial time by restricting what reasoning can be used.
Such a specialized form of reasoning is what one of the earliest reasoning
modules that Cyc implements. Cyc has found such restrictions as both
useful and desirable for such problems. (06)
When a question is asked at Cyc's default query interface, there are a large
number of settings that one may set. Cyc's default for classifying some
item is exactly this module that OWL implemented over half a decade later.
This module will not look for chains of rules to conclude the classification. (07)
Cyc encourages rules that conclude a classification to be forward chaining,
so that this simple module can be used for classification purposes on
queries. One can change the settings on the query interface and allow for
additional reasoning (with optional time, depth, and number of answer
cut-offs). (08)
- doug foxvog (09)
...
> The issues I want to emphasize are the ones in Tim's original proposal.
>
> John (010)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (011)
|