The Hundredth Monkey Phenomenon

About two months ago, Nova Spivack wrote a post entited, The Ontology Integration Problem in which he discusses the problems he’s experienced in developing ontologies for use in his startup company, Radar Networks. The crux of his article is:

In my own experience designing a number of OWL ontologies (500 classes - 3000 classes on average) it has often been easier to create my own custom ontology branches to cover various concepts than to try to integrate other ontologies of those concepts into my own.

Having written several ontologies during my tenure at Big Blue, I do sympathize with his complaints. Unless you’re writing an application for which there is an existing ontology that fits your application’s specific needs, it is often much easier to write your own OWL document than to mix and match vocabulary from various other ontologies. For example, we wrote a semantic web rich client application in Java and SWT for The Center for the Development of a Virtual Tumor (CViT) in which we had to represent users in the system. A user object instance is associated with an account instance, and the user, depending on his type–researcher, team leader, lawyer, etc.–has various properties. Some properties are common to all user types, and most of those exist in the FOAF vocabulary. I ended up extending the foaf:Person class in our application’s user/account ontology and found that this worked out well for us. However, I can’t imagine that I am the first person in the world to create a “Researcher” or “Lawyer” type in an ontology, or especially the venerable “Account” type. Where are the others? How do I find them? Are they well-designed ontologies or just quickly hacked-up OWL documents?

Fortunately, certain communities, especially in life sciences, have several large efforts in progress to standardize ontologies for common and important areas. For example, there is the Scientific Publishing Task Force at the W3 whose goal is:

To develop a general purpose ontology for self-publishing single experiment in RDF format that will facilitate data sharing, discovery and integration. Applications such as web-publishing tool and semantic search engine that are built on top of this ontology will demonstrate the emerging semantic standards and technologies can help developing more interactive scientific communities centered around user-generated scientific contents on the web .

As Nova mentioned, the great value of the semantic web is that it allows automagic integration of data from all over the place, and the only way that we’re going to acheive this is if the semantic silos agree on vocabulary–or at least on a way to easily map between different-but-similar vocabularies. Because of this, I think that in each domain there will be natural convergence on whatever ontologies are needed by the domain. Furthermore, cross-domain ontologies will help specific ontologies have some common ground. If, for example, Nova’s company used foaf:Person as the basis for users in their architecture, even if he might have created his own specialized vocabulary in addition to the inherited properties, my system would at the very least be able to query his system for the subset of properties that we both have in common.
Even though FOAF’s certainly not perfect or complete, it’s commonly used, which makes it valuable (of course, there are arguments against using an ontology just because it’s popular).

So while I agree that, at the moment, there are only a handful of commonly used, useful OWL ontologies that one can draw on, we’re still early in the age of the semantic web, and this problem will only improve with time and with efforts to develop the types of reasonable, common ontologies that Nova, and I, and anyone else working on this type of stuff, will really need in order to allow true data integration of our semantic silos.