Last Friday, while I was discussing ontologies and decisions that need to be made in ontology development with some work colleagues, one of the phrases that cropped up more than once is “be sensible”. Being sensible isn’t always as easy as it seems, but one way to be sensible is to choose an ontology development methodology and make use of before you even write down your first ontology class name. If you want lots of people to use an ontology, you need to involve at least some of those people in its development.
As a timely accompaniment to this thought, in the past week Frank Gibson has published a pre-print version of a methodology for distributed ontology development called Developing ontologies in decentralised settings (by Alexander Garcia, Kieran O’Neill, Leyla J. Garcia, Phillip Lord, Robert Stevens, Oscar Corcho, & Frank Gibson).
While Frank himself has referred to it as “dry”, I think that does it a disservice (but perhaps I’m biased because I know him and also because I like methodologies and standards!). This paper would better be described as comprehensive. I’d like to cover a few sections of the paper that I found the most interesting, to whet your appetite for reading the whole thing.
Firstly, Garcia et al. mention one overriding focus of the bio-ontology community: ontology development without any accompanying ontology development methodology:
‘The research focus for the bio-ontology community to date has typically centred on the development of domain specific ontologies for particular applications, as opposed to the actual “how to” of building the ontology or the “materials and methods”[…] This has resulted in a proliferation of bio-ontologies, developed in different ways, often presenting overlap in terminology or application domain.’
Both in programming and in ontology development, I find it very hard not to head straight for working on the “interesting” bits without thinking through the best way to go about it. However, even though I find it difficult to follow a particular methodology, the benefits outweigh the downsides.
Garcia et al also list a kind of minimal set of requirements for an ontology methodology:
‘A general purpose methodology should aim to provide ontology engineers with a sufficient perspective of the stages of the development process and the components of the ontology life cycle, and account for community development. In addition, detailed examples of use should be included for those stages, outcomes, deliverables, methods and techniques; all of which form part of the ontology life cycle.’
So far, these are useful statements for anyone building an ontology, but this paper concentrates on distributed ontology development, and presents Melting Point (MP), an ontology methodology specifically designed for distributed, community-driven ontology development. It was created as a “convergence of existing methodologies, with the addition of new aspects” as “no methodology completely satisfies all the criteria for collaborative development” (pg. 2). A useful overview of MP is available from Figure 3 in the paper, which describes the life cycle of the MP methodology including its processes and activities.
This paper has a thorough review of nine existing ontology and knowledge engineering methodologies (see Table 1 and Section 4.2 particularly), and clearly explains why MP was important to develop. I encourage anyone interested in building ontologies to read this paper for its background information, and especially encourage anyone interested in distributed, community-driven development of ontologies to read this and determine if MP might be the right methodology for you.
I’ll finish as Garcia et al. has, with their concluding paragraph. Enjoy!
‘As we increasingly build large ontologies against complex domain knowledge in a community and collaborative manner there is an identified need for a methodology to provide a framework for this process. A shared methodology tailored for the decentralized development environment, facilitated by the internet should increasingly enable and encourage the development of ontologies fit for purpose. The Melting point methodology provides this framework which should enable the ontology community to cope with the escalating demands for scalability and repeatability in the representation of community derived knowledge bases, such as those in biomedicine and the semantic web.’