John, (01)
There is a certain amount of doubt in the businesses I work for as to
whether ontologies are anything more than hype. The approach I take is to
say the following: (02)
1) An ontology is a type of data model. It differs from a conventional data
model in that it is written in logic. (03)
2) In conventional data models, the model is embodied as special purpose
code that has to be created specifically for the model. With an ontology,
one uses a general purpose reasoner that works directly from the logic
representation. (04)
3) If you want to interoperate between two systems with different data
models, you need to create a mapping between them. With conventional models,
this mapping must be turned into special purpose mapping code, whereas with
an ontology, the mapping is also written in logic, and so the same general
purpose reasoner can be used. (05)
4) The advantage of ontologies is therefore that the model and the mappings
are data for a general purpose reasoner, rather than requiring the
generation of special purpose code. This makes it easier to maintain
interfaces, since the updates can be sent out as data rather than code
patches. (Assume here that there may be many hundreds of systems that carry
the interface). (06)
The business advantage of an ontology over a conventional data model is
therefore likely to be in the costs of maintaining interfaces. (07)
Note: The biggest cost of an ontology is confirming its grounding - that is,
confirming that the data means what you think it means. A failure to do that
may mean that the systems may need to be closed down for two or three days
to recover them to a safe state - for example, in one data exchange,
importing bad data could have stopped 5,000 people working for two or three
days (say £2-3 million per day). Since most of the data I deal with is high
value, there is a high risk if you get it wrong - which means working closed
worlds so that you can trust the data sources. (08)
The question is not, "Why do we need ontologies", since there are plenty of
ways to do the same thing without ontologies. Rather, the question is, can
we do what we want to do in a cheaper, more reliable way? (09)
The biggest challenge is to make ontologies interesting, in the sense that
they reveal the ideas driving their usage. In the teaching of mathematics,
at least at the higher levels, the function of proof is to reveal the
mathematical ideas used, rather than merely confirm a fact. At the moment, I
don't know how to do that in an ontology - it was bad enough in formal
specifications. Rather, we are looking at how to do the things we want in a
conventional (though not very conventional) data model, and once we know how
they work, transforming them to an ontology. (010)
Sean Barker
Bristol, uk (011)
_________________________________________________________________
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 (012)
|