background image
he Semantic Web as it is advocated by
people like Tim Berners Lee and James
Hendler does not take enough advantage
of the experience built up in knowledge engineering
and conceptual modelling.There is this anarchistic
idea of the Web as a place where everyone can do his
or her own thing. I have no problem with that; a lot
of people are able to find what they want on the
Web. But if you want real interoperability, with search
engines that can grasp the intended meaning of infor-
mation, that approach falls short.To create a real
Semantic Web we have to develop and use well-
founded generic ontologies, based on linguistics
and logics.'
Nicola Guarino has clear views on the Semantic
Web and its development. He is a senior researcher at
the Institute for Cognitive Sciences and Technologies
in Italy, where he leads the Laboratory for Applied
Ontology. Since 1991 he has played an active role in
the Artificial Intelligence community in promoting
the interdisciplinary study of ontological foundations
of knowledge engineering and conceptual modelling.
Guarino: `In our Laboratory the focus is on content
and not so much on representation.The use of
ontologies is unavoidable when referring to content.
People do it implicitly all the time when they are
communicating and trying to understand each other.
If we want machines to understand each other, in
other words real interoperability, we need to make
these ontologies explicit in an unambiguous way.'
An ontology is a hierarchical description of the
relations between concepts in a certain domain plus
an unambiguous description of the concepts
themselves. As they are created for a certain domain,
ontologies often fail to be interoperable, because of
the ambiguity that results from the use of the same
terms for different concepts (and vice versa) between
different domains.The term `net' for instance has
quite a different meaning for Web designers and
fishermen.That is why there is a need for well-
founded generic ontologies. An example of a generic
ontology is the term `part', which can have different
meanings both within a domain and between
domains. For instance, the violist plays a part in the
orchestra. His finger is part of him. Can his finger be
part of the orchestra? According to Guarino, this is a
genuine ontological problem that can only be solved
by giving an unambiguous meaning to the term `part'.
Another example, cited by Guarino, is the term
`in'.What exactly are you describing when you say
the spoon is in the cup? Does it mean that the spoon
is totally embedded in the cup or is it only partly in
the cup? Guarino: `These examples seem trivial, but
if you want real interoperability between different
knowledge domains you will have to prevent the
problems that come with the ambiguity of day-to-
day language.'
In this respect Guarino thinks it is a drawback that
computer science curricula scarcely ever contain an
introduction to ontological foundations of conceptual
modelling. `Students learn all about Java, HTML and
C++ and name all the other languages, and they also
learn how to use these. But when they graduate they
hardly know a thing about formal ontology. I really
think people should know more about the work on
ontology that has been done in philosophy. It is
certainly not much harder to acquire than, say,
studying differential equations, or learning how
to use Java.'
It seems as if it is an enormous job to develop
well-founded generic ontologies, but it is not as
enormous a task as it appears. Guarino: `I would say
that a few dozen would get you on the way nicely.
But you have to take the fundamental route. At the
moment development of the Semantic Web is driven
by the need for short-term results. Hence,
interoperability is realised by putting the right tags
on the information.That is not what I call semantics;
that is syntax. XML and RDF are very useful for this,
but they fall short when you want to create a real
Semantic Web.'
Laboratory of Applied Ontology,
DigiCULT 25
, I
By Joost van Kasteren