Building ontologies with basic formal ontology* (Bobfo) is a 200 page introduction to the subject by Robert Arp, Barry Smith and Andrew Spear.* While ontology in its broad sense, the theory of what exists, is all-encompassing and philosophical in scope, Bobfo focuses on analyzing the ‘information domain’ with an intended application in IT and data modeling. The ontologist is therefore a journeyman modeler who can identify and extract the essence of data, relationships and indeed, everything else in a field of activity. In the case of Bobfo’s authors this is medicine and the biological sciences, but the approach is intended to have application everywhere.
To understand the basic formal ontology itself we would recommend viewing co-author Smith’s video which succinctly explains the failure of the semantic web, linked open data (LOD) and ontology to date. Smith categorizes the much-vaunted constellation of semantic LOD as a ‘hairball’ from which information can only be extracted with considerable manual effort. The BFO sets out to fix linked data’s ‘anarchy and chaos’ with domain-neutral standards for building ontologies, shared by all.
Bobfo defines an ontology as follows, ‘a representational artefact, comprising a taxonomy as a proper part, whose representations are intended to designate some combination of universals, defined classes and certain relations between them.’ A taxonomy is then defined as a simple hierarchy while ‘universals,’ and synonyms ‘classes,’ and ‘types’ are defined as groups of the entities in the world that is being described.
As you will have gathered this is a pretty dense oeuvre as perhaps befits a field so close to philosophy. One page 6 we are plunged into a discussion on terminology research, on how the ‘concept orientation’ of ISO, the international standards association was derived from the ‘phenomenonalist’ ideas of the Vienna Circle. This approach has been replaced by the ‘realist orientation.’ The realists downplay the ‘ideas in people’s heads’ to focus exclusively on ‘labels that represent entities in reality.’
Following two chapters on best principles for ontology design, the book gets down to business with the BFO itself. The BFO is a small, top-level (a.k.a. upper level) ontology designed to support data integration in scientific research. It addresses the time-dependent nature of measurement by distinguishing between a continuant and an occurrence. Other BFO concepts include role, disposition, boundary, spatial region and relation. These get rather thorough treatment, with erudite and interesting asides such as a digression on Arthur Eddington’s two tables.
The rubber hits the road in chapter 8 describing the BFO at work. Concretely this means the use of the web ontology language OWL, the W3C’s resource description framework RDF and the ontologist’s favorite tool of the trade Protégé. A short section on ‘facilitating interoperability’ outlines the potential benefits, with a pointer to the work of the Open biological foundry, an ‘expanding virtual framework for navigating massive amounts of biological and clinical data.’
BFO’s users are mostly in biosciences as shown on the Ifomis website. In our quick spin through the list we found ontologies for email, economics and petrochemicals. But not all appear to be maintained, far from it. One noteworthy use case is the USGS that has leveraged the BFO in its surface water ontology.
Bobfo underscores the weakness of current domain-specific attempts at semantics and enumerates many of the modelers’ pitfalls. In the end, the success or otherwise of the BFO approach depends on whether the benefits that accrue from decomposing a specific domain into its ontological components exceed the considerable intellectual effort that this requires. Those interested in such matters may like to sign up for the upcoming International conference on formal ontology in information systems in the beautiful town of Annecy, France next July.
* By Robert Arp, Barry Smith and Andrew Spear. 2015 MIT Press. ISBN 978-0-262-52781-1.
© Oil IT Journal - all rights reserved.