More ontology design parameters and dependencies between them

Following up on the MTSR’09 short paper on ontology design parameters [1] for agri-ontologies (blogged about earlier), I received an invitation to write an extended version of it for the International Journal on Metadata, Semantics and Ontologies. This much extended version—20 pages double column vs the earlier 6 page LNCS-format—got accepted recently and is currently in print [2]. Reading the small print of the author agreement, I seem not to be allowed to put either the abstract or the paper online (but clearly, emailing the preprint to interested people does not constitute publishing it), so here goes an informal rendering of the abstract.

Anyone in the ontologies arena is well-aware of the fact that the development, adoption, extension, and use of ontologies is increasing, yet at the same time such efforts are hampered by the new challenges wider uptake generates, such as determining which ontologies to reuse, if any, and which language to use. Several ontology development methodologies are available, such as Methontology and the emerging NeOn methodology, which provide scenarios but do not yet address the dependencies between the permutations at the different stages in the development process. In an attempt to improve the efficiency and effectiveness of this endeavour, I have grouped the myriad of inputs into types of parameters and examined the dependencies between them. The types of parameters that were considered are: purpose(s) of the ontology, reuse of different types of ontologies, different ways for bottom-up ontology development, ontology languages (mainly the DL-based OWL species and several proposed extensions), and reasoning services (grouped into four types).

Subsequently, all dependencies between these parameters have been assessed on their feasibility of combination and useful combinations are motivated (e.g., why ontologising thesauri with OWL 2 DL and further extensions may not be such a good idea, but representing a scientific theory is, and using OntoClean with a bare taxonomy is also a good idea, but not DOLCE with Ontology-Based Data Access). The dependencies between the assessed parameters are due to, primarily, computational challenges and types of subject domain of the ontologies (motivated more extensively in the paper). In addition to the review and theoretical assessment, the analysis was assessed against a set of randomly selected ontologies and a survey among ontology developers; the results obtained concur with the theoretical assessment.

At some point in the future, it could be nice to have this integrated in software-supported ontology development environments (ODEs) as some sort of decision procedure. At least it will be able to suggest sensible combinations and discourage upfront combinations that will be tedious and overly resource-consuming to implement, thereby making the ontology engineering process more easily accessible and efficient by proposing combinations that lead to feasible implementation scenarios.

References

[1] Keet, C.M. Ontology design parameters for aligning agri-informatics with the Semantic Web. 3rd International Conference on Metadata and Semantics (MTSR’09) — Special Track on Agriculture, Food & Environment, Oct 1-2 2009 Milan, Italy. Springer CCIS 46, 239-244.

[2] Keet, C.M. Dependencies between ontology design parameters. International Journal on Metadata, Semantics and Ontologies. In print.

Advertisement