Forwarded message 1
Kostas,
There is a large (and growing larger) literature out there. In general
you cannot do automated mapping, since someone or something needs to
know the semantics of both ontologies. State of the art is
semi-automated. People try to get by with weak methods, i.e.,
string/substring identity, graph homomorphism, etc., but these are
really insufficient. The best method is to show that two ontologies have
equivalent formal models (or subsets, overlaps, etc.), but the problem
is that that is easier said than done. Usually the ontologies are
expressed in different representation languages (possibly with
ill-defined formal model theories). Also, it may be the case that you
need to have as much semantics preservation as possible as you go
between a semantically well-defined ontology/taxonomy and an ill-defined
ontology/taxonomy, the latter of which may have "pockets" of
well-defined semantics or has [typically unsound] reasoning methods
which nevertheless may be pragmatically and locally useful: many of the
procedural and quasi-declarative reasoning systems and applications in
the world are unsound.
The question of semantic mapping is being addressed in 3 technical areas
that I am aware of: database, thesaurus, and ontology communities. On
the one hand the task is harder in ontologies because of the semantic
richness, on the other it's easier because of the typically more
precisely defined semantics. Also, the ontology community has mostly
focused on merging ontologies, not mapping ontologies, but in many cases
you need to preserve the independence of the ontologies (possibly
different owners, standards, etc.) and hence just map.
I am very interested in this area myself. We recently had a paper in
F.OIS-01 "Ontological Engineering for B2B E-Commerce" where we
illustrate somewhat the problem of semantic mapping, from the
perspective of B2B.
Here is just a short list of approaches and related literature (yes,
formalization of context is very close to semantic mapping), don't have
the exact citations right now:
Microtheories: Lena, Guha, et al, 1990, etc., Cycorp
“Little Theories” and Theory Interpretation: Farmer et al, 1994, MITRE
Articulation Ontologies: Wiederhold, Mitra, Jannink, 2000, Stanford U.
Graph Homomorphisms: Many
Conceptual Anchoring, etc.: Noy, 2001, SMI
Local Models Semantics (Context): Giunchiglia, Ghidini, 1997, U. Trento
Formalized Context: McCarthy, Guha, Buvac, 1990, etc. Stanford U.
Morphisms (Category Theory): Many
Information Flow Theory: Barwise & Seligman, 1997
Information Flow Framework Candidate Upper Ontology (IEEE Standard Upper
Ontology): Robert Kent, 2001
Intercontext Correlation: Skvortsov, Kalinichenko, 2001, Institute for
Problems of Informatics, Russian Academy of Science
Schema Mapping: Rahm & Bernstein, Universität Leipzig, 2001
Ontolingua/Chimaera, Fikes & McGuinness, 1999, etc., Stanford U.
Ontomorph, Chalupsky, 2000, ISI.
Also, of course, there is some work being done on approximating semantic
equivalence statistically, have to look for references.
Hope this helps some.
Leo
> Kostas Kastradas wrote:
>
> Hi,
> I am looking for algorithms that map, automatically not manually, two
> ontologies. First, I want to understand the mechanisms of mapping amd
> second I want to apply them to map two simple, not large scale,
> ontologies.Any suggestions?
> P.S.:I am aware of the Anchor algorithm that it was suggested by
> Natasha F. Noy but I would like to see some more just to have a
> general idea of mapping
> Thanks
> Kostas Kastradas
--
_____________________________________________
Dr. Leo Obrst The MITRE Corporation
mailto:lobrst@mitre.org Intelligent Information Management/Exploitation
Voice: 703-883-6770 7515 Colshire Drive, M/S W640
Fax: 703-883-1379 McLean, VA 22102-7508, USA
----------------------------------------------------------------------
To unsubscribe, send email to majordomo@smi.stanford.edu with
"unsubscribe protege-discussion" in the message body (no quotes).
If this doesn't work, contact owner-protege-discussion@smi.stanford.edu