Semantics or not semantics?

I have seen the following arguments proposed as to why it might be a
good idea NOT to have a clear/formal semantics for a web based
ontology language (I paraphrase):

a) The web is bound to contain many doubtful and conflicting
   assertions.

b) Semantics mean inference, and it is impossible to enforce/guarantee
   that inferences will be made.

As far as (a) is concerned, being unclear as to the intended meaning
of such assertions only exacerbates the problem as we first have to
"guess" a (possibly incorrect) meaning for an assertion before deciding
on the degree of scepticism with which to treat it. As for (b), a
clear semantics only allows us to determine when a given inference is
valid with respect to a set of assertions and says nothing about
whether the inference can/should be made.

In both cases, semantics provide the necessary "yardstick" against
which we can measure - they say nothing about how we should deal with
the resulting measurements.


The need for a clear semantics for DAML-ONT has been well made by
Peter Patel-Schneider. A great deal of time has already been wasted
arguing about what the specification means when we should have been
discussing the design of the language. This is in contrast to OIL,
which has always had the benefit of a denotational semantics, and
where fruitful discussion has been possible and has led to the
correction of errors and infelicities in the language design.

A good example is the great "equivalent" debate. In OIL, the meaning
of a class is given by the set of objects in the world that it
denotes. Two classes (either names or complex specifications) are
equivalent if they denote the same set of object. Simple, clear and
unambiguous.

Received on Monday, 16 October 2000 09:11:14 UTC