Pre-coordination

[Norman Gray]

Greetings, all.

At the semantics session in Naples today, there was a brief discussion on the relationships between vocabularies and ontologies.  I mentioned that there was a semi-standard pattern for this, called 'pre-coordination'.

The idea is that you devise an ontology for your problem domain, which might be quite sophisticated and expressive, but inappropriate for distribution.  You can then produce a pre-coordinated version (in a step which is analogous to a compilation step) as a very simple ontology or vocabulary, which can be used in much simpler tools.

See:

http://www.debugit.eu/events/documents/Schober%20PrePostCoord_Lissabon.pdf A presentation on "Requirements for pre- and post-coordinated ontology engineering"

Example:

Pneumonia	≡	Inflammation ⊓	has-participant. LungTissue
AcutePneumonia ≡ Pneumonia ⊓ bearer-of. AcuteQuality
BacterialPneumonia ≡	Pneumonia	⊓	has-agent. BacteriaPopulation
ViralPneumonia	≡	Pneumonia	⊓	has-agent. VirusPopulation

Thus, the ontology has classes 'Inflammation' and so on, and properties like 'hasAgent', and the pre-coordination step generates the terms Pneumonia, AcutePneumonia, and so on, from this.  The latter terms are more immediately usable (in this domain, it tends to be medical billing applications), but would be a royal pain to maintain in that form, because there's thousands of them.

There's a likely-looking review article at doi:10.1108/00220410910983128, but my library doesn't have access to that journal, and I can't find a preprint.

In the astronomical case, one could imagine generating EmissionH2, EmissionO2 and so on, by pre-coordination from classes of 'Emission' and a collection of elements.

Pre-coordination is a pattern rather than an implementation, but it illustrates how one might address the problem that Gerard raised.

All the best,

Norman


-- 
Norman Gray  :  http://nxg.me.uk
School of Physics and Astronomy, University of Glasgow, UK