The big question that surrounds Demystifying OWL for the Enterprise* (DOE), a new book by Semantic Arts’ Michael Uschold, is, is it timely or too late**! The publishing blurb has it that semantic web technology stack has seen a ‘slow incubation period of nearly 15 years’ but that today, ‘a large and growing number of organizations’ now have one or more semweb projects underway. Oil IT Journal has been tracking semantic technology oil and gas since 2003 and has reported, in over 200 articles, on progress that has been faltering to say the least. Semantic technologies still attract enough R&D cash to keep students on either side of the pond in pot noodles, but industry at large would appear to be turning more to commercial ‘graph database’ tools in the (so far few) occasions where there is a pressing need to use something other than a relational database.
Notwithstanding these reserves, ontologies are key to information systems of all sorts and those embarking on any kind of natural language processing venture are likely to come across the arcane terminology of the field. Perhaps a better question to ask of DOE is, does it do a good job of demystification? For this non-specialist but curious reviewer the answer is yes. DOE focuses on ‘the 30% of OWL that gets used 90% of the time’. Moreover, this 30% can be ingested in Part 1 (74 pages). So, what is an ontology anyway? It is a model that represents some subject matter. And what is OWL? The claim is that OWL provides a schema for a particular domain that is more informative and flexible than the RDBMS schema. OWL sees the world as triples, subjects and objects connected by relationships. Google is an InstanceOf a Corporation. And a Corporation is AKindOf Legal Entity. From which a computer can figure that Google is a Legal Entity.
DOE goes on to explain namespaces which allow different ontologies to co-habit, and resource identifiers, which pin down objects and relationships into their namespace. The use of the term ‘uniform resource identifier’ URI is confusingly close to URL, especially as they can be one and the same. But DOE does a good job of untangling these. We dug deeper into DOE to seek enlightenment on another term that has always puzzled, ‘reification’. This is explained as a kind of work-around to allow for the mapping of one-to-many relationships in OWL. While this is clear enough, this intrinsic awkwardness does little to convince the reader that OWL has the upper hand over the RDBMS.
Where OWL has a good claim to primacy over the relational model is in the field of machine reasoning. This is achieved by applying logic to syllogisms – if all men are mortal and Socrates is a man then Socrates is mortal. That’s simple enough, but what follows is where the going gets hard. As hard, in fact, as formal logic. The assumption is that the computer will be able to ‘reason’ across a large set of data such that, again with the healthcare example, the reasoner will be able to figure ‘who are all the patients that Jane has given care to’. And this without any such direct relationship being built into the model. It is not hard to see how this could be extended to other fields such as geology with multiple overlapping categories.
Part 2 concerns the 10% you don’t use all the time. One field that we have covered in the past is the Upper Ontology, an overarching ontology that can tie sub-ontologies together. DOE recommends Gist Semantic Arts own ‘minimalist’ UO with a nod to BFO, the basic formal ontology that is said to be ‘widely used by scientists’. The short section on the UE fails to capture the heated debate surrounding these issues although in his conclusion, Uschold observes that ‘if there are two ontologists in the room there will be at least three opinions on how to model a given thing’.
DOE does a good job at demystifying OWL. But is the technology really all it is cracked up to be? Chapter 7 handled this question asking by enumerating OWL’s limitations. This chapter alone merits the book’s purchase by any contemplating an OWL project. The discussion of properties as property values shows how hard it is to model engineering units in semantic triples, suggesting some workarounds for such awkwardness.
* Morgan & Claypool’s series on the Semantic Web: Theory and Technology: Demystifying OWL for the Enterprise, Semantic Arts, Inc. Paperback ISBN: 9781681731278 eBook ISBN: 9781681731285 Hardcover ISBN: 9781681732831 May 2018, 264 pages .
** Tim Berners-Lee unveiled the embryonic semantic web at the first meeting of the W3C in 1994. OWL was introduced in 2004.
© Oil IT Journal - all rights reserved.