[Fwd: Ontology mapping] Re: model-model mapping

3rd message from Protege list. 
-- 
_____________________________________________
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

Forwarded message 1

  • From: Natasha F. Noy <noy@SMI.Stanford.EDU>
  • Date: Mon, 26 Nov 2001 16:58:40 -0800
  • Subject: Re: Ontology mapping
  • To: protege-discussion@SMI.Stanford.EDU
  • Message-Id: <4.2.2.20011126163550.00c131b0@smi.stanford.edu>
Hi,

As people on this list have pointed out, you won't find many algorithms for 
mapping between ontologies completely automatically. And, in any case, you 
need to be aware that all of those algorithms are (collections of) 
heuristics. Don't get me wrong -- there is nothing wrong with the 
heuristic-based approach. In fact, I personally believe that this approach 
is the only possible one in the case of ontology mapping: So much of the 
semantics is always implicit (even in ontologies), that it is almost 
impossible to have any formal definition of what it means for a concept in 
one ontology to map into concept in another ontology. If you want this 
process to be precise, you always need some form of human involvement, if 
only to verify the mappings with which the automatic algorithm came up with.

That being said, recently there has been a number of approaches for finding 
candidates for mapping automatically, both in the ontology-mapping/merging 
field and in the schema matching field. In case you haven't got enough 
references already, here are some more for your list:

The ONION project in the Stanford DB group, in particular 
http://www-db.stanford.edu/SKC/publications/match.ps They use some seed 
matching rules from the user to generate new matching rules (and have the 
user prune them).

FCA-merge from University of Karlsruhe: 
http://www.fzi.de/wim/publications/2001/federated_ijcai_ws.pdf and their 
IJCAI-2001 paper. They use a set of instances that are common to the two 
ontologies to generate some possible matches (assuming these common 
instances exist).

The LSD project at University of Washington 
(http://www.cs.washington.edu/homes/anhai/lsd/lsd.html) uses 
machine-learning techniques to use XML data to produce matches.

Our Anchor-PROMPT algorithm 
(http://smi-web.stanford.edu/pubs/SMI_Abstracts/SMI-2001-0889.html) is 
another one that attempts to produce some candidate matches between 
ontologies for user's perusal.

For a review of automatic and semi-automatic mapping for database schemas, 
check out the following very comprehensive review: 
http://research.microsoft.com/scripts/pubs/view.asp?TR_ID=MSR-TR-2001-17

For a couple of recent (not included in that review) semi-automatic 
approaches to schema matching see the similarity-flooding algorithm 
(http://dbpubs.stanford.edu:8090/pub/2001-25) and the Cupid project 
(http://research.microsoft.com/~philbe/CupidVLDB01.pdf)

Enjoy,

Natasha



----------------------------------------------------------------------
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

Received on Friday, 18 January 2002 18:46:13 UTC