Mapping between different knowledge sources

Apologies in advance as this is skipping on from the use cases, but it seems
to me one of the key questions in SIMILE (and the semantic web) is whether
it is possible to interoperate between multiple vocabularies. Interestingly
quite a bit of work has been done on this before - as I have mentioned the
KRAFT project between BT PLC and Liverpool, Aberdeen and Cardiff
Universities considered this problem.
http://www.csd.abdn.ac.uk/~apreece/Research/KRAFT.html
KRAFT stands for Knowledge Re-use and Fusion / Transformation

This section from "Ontological Structures for Knowledge Sharing" M J R
Shave, 
www.csc.liv.ac.uk/~mshave/NRIN+.ps
<http://www.csc.liv.ac.uk/~mshave/NRIN+.ps
seems particularly relevant:

"The specification of mapping functions is clearly a key task in the
construction of a KRAFT network. Knowledge sources can differ in their
content (their data and its structure), their paradigm (the modelling
convention used, such as an object oriented database or a knowledge base),
their representational language (such as predicate calculus, or frames), and
their ontology. 

The resolution of each of these differences presents problems, but
ontological mapping is in many ways the hardest task, if only because it is
the least well understood. Considerable attention has been devoted in the
KRAFT project to identifying and classifying the ways in which differences
between ontologies can occur, and how such mismatches can be resolved. In
the simplest cases differences can be caused by synonyms (`client' and
`customer' often have the same meaning), by homonyms (a `crest' can be a
badge, or the top of a wave or hill), or by differences in underlying
assumptions (`pass' and `fail' are familiar concepts but have widely varying
interpretations). Other mismatches are more complex and less straightforward
to resolve. 

Two broad categories of mismatch have been recognised. Conceptualisation
mismatches result from differences in the categories or data structures
used. Explication mismatches result from the definitions used, which may
differ in their terminology, their formulae, or the concepts which they are
defining. Some examples are: 

Ontology 1 uses the classes mammals, birds 
Ontology 2 uses the classes carnivores, herbivores 
A class conceptualisation mismatch 

Ontology 1 uses the relation hascomponent 
Ontology 2 uses the relation ismadeof 
[eg The pair {house, brick} meets either ontology, but {house,roof} fits
only the first] 
A relation conceptualisation mismatch 

Ontology 1 : ship (X) < seagoing (X) ^ large (X) 
Ontology 2 : vessel (X) < seagoing (X) ^ large (X) 
A term explication mismatch 

Ontology 1 : ship (X) < seagoing (X) ^ large (X) 
Ontology 2 : vessel (X) < floating (X) ^ big (X) 
A term and formula explication mismatch 

Ontology 1 : ship (X) < seagoing (X) ^ large (X) 
Ontology 2 : whale (X) < seagoing (X) ^ large (X) 
A concept and term explication mismatch 
A detailed discussion of these issues can be found in [8]. 

[8] P R S Visser, D M Jones, T J M Bench-Capon and M J R Shave, "An Analysis
of Ontology Mistmatches: Heterogeneity versus Interoperability", AAAI Spring
Symposium of Ontological Engineering, California 1997

br

Mark H. Butler, PhD
Research Scientist                HP Labs Bristol
mark-h_butler@hp.com
Internet: http://www-uk.hpl.hp.com/people/marbut/

Received on Monday, 24 March 2003 07:43:09 UTC