Pat and Rob, (01)
I'll start with Rob's note because it leads to some issues
that are relevant to a Foundational Ontology. (02)
RF> In large part my posts come down to encouraging exactly this
> approach: that the best way to model knowledge is by directly
> processing patterns found in raw text. (03)
I agree that is an excellent technique for processing a great deal
of NL text, but it's not the only tool I'd want in my toolkit. (04)
RF> Have you at least moved away from the position that some kind
> of abstract meaning representation is necessary for "meaning-
> related" language processing problems like speech recognition? (05)
For at least the initial stages of speech recognition, the
pattern-based approaches are very important. But once you have
an idea about what is being said, it's necessary to get into
the details of the subject matter for deeper understanding. (06)
"Meaning" depends critically on what "language game" is being played.
If you're talking about a chess problem, the system will have to
relate the language to chess positions and tactics. (07)
If you're talking about hiking, the system will relate the
language to the territory over which you're hiking. (08)
In every application or language game, the same extra-linguistic
background needed for understanding and reasoning about the
subject will be necessary for understanding and generating
language about that subject. (09)
There is no such thing as a one-size-fits-all ontology that
can cover chess, hiking, programming, eating, cooking, chemistry,
surgery, driving a car, and football. For most of those subjects,
a very large part of the background knowledge will *not* be
expressed in either verbal patterns or some version of logic. (010)
PC> The need for an FO arises from the need for interoperability. (011)
Just consider that list of activities mentioned above. People
don't interoperate with equal facility on all those areas.
The number of professional football players who are also
expert chefs, master chess players, and surgeons you would
trust for a triple bypass operation is rather small. (012)
The idea that a computer system requires equal semantic coverage
of a wide range of areas in order to achieve interoperability
has no basis in reality. Just look at a Swiss Army Knife.
It's not bad for a quick fix in many cases, but for each
application, it's better to use a more specialized tool,
if you have one. (013)
John (014)
_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (015)
|