Key questions to address in choosing an ontology:
- Is it a good fit with my research questions?
- Has it been used by other users in my research community?
- How easy is it to find learning resources for this ontology?
- What is the long term support of the ontology itself (sustainability)?
Aside from choosing an ontology that will help answer a project’s particular research questions, considerations of the level of adoption of the ontology in the research community should be considered in your aim to create broader data integration in the long run. Owing to the logical construction of ontologies and the subtleties this can entail, practical educational material to learn and be able to use the main concepts of the ontology is essential.
It is important to emphasize that an ontology is meant to be implemented by the researcher with the aid of a computer scientist in order to accurately represent their . The researcher must take responsibility for modelling their own data and make informed decisions on the implementation and its consequences. The researchers knows their data best and therefore also how it should be semantically represented. This is a radical shift from the notion of data management where data modelling is consigned to the computer scientist alone who, for obvious reasons, has no necessary knowledge of the field of research being modelled. For this reason it is important that the ontology be learnable by the intended users even if, in everyday research, they will not have to make reference to the ontology (normally hidden by simplified data entry forms or well constructed intuitive query systems).
Finally, it is important to see that the community that develops and maintains the ontology is active and responsive. Ontologies, as with any standard, are living systems that require maintenance and updating to respond to new challenges and unforeseen problems. This higher level modelling work is time-consuming both in conceptualization and documentation. For most projects, taking on the task of ontological modelling itself (except in a limited scope) would be unfeasible, but should rather rely on the groundwork of the expert community designing their adopted ontology.