I’d pretty much line up intrinsic with semantics, and extrinsic with pragmatics, yes. Steven R. Ray, Ph.D. Distinguished Research Fellow Carnegie Mellon University NASA Research Park Building 23 (MS 23-11) P.O. Box 1
Moffett Field, CA 94305-0001 Email: steve.ray@xxxxxxxxxx Phone: (650) 587-3780 Cell: (202) 316-6481 Skype: steverayconsulting 
From: Obrst, Leo J. [mailto:lobrst@xxxxxxxxx]
Sent: Tuesday, December 18, 2012 2:42 PM
To: steve.ray@xxxxxxxxxx; Ontology Summit 2013 discussion
Subject: RE: [ontology-summit] Scope of ontology: Issues: Well, I mostly agree, given that we have kind of made this distinction by fiat, but since I think even intrinsic ontology evaluation is not just syntactic (well-formedness) but semantic (e.g., using ontological analysis of the taxonomic backbone along the lines of OntoClean). The demarcation is between the semantics and pragmatics, I think, and so, if by extrinsic we mean pragmatics, I am fine with that. Is that what we mean? I don’t think this is an extraneous conversation we are having, but in fact goes to the points of the Ontology Summit, which we are trying to work out. I was responding originally to Alan’s message, which we had not yet pigeon-holed as being A Intrinsic or B Extrinsic or C. We may choose to place our notions in bin A or B or C, but the notions are important. Thanks, Leo I think the conversation is veering off point. The point being discussed (whether you need to have an application in mind when evaluating an ontology) goes back to the “intrinsic” versus “extrinsic” evaluation. One can evaluate an ontology on an intrinsic basis, much as you could evaluate a sentence: Is it well formed? Does it conform to proper rules of grammar? For an ontology, questions would relate to proper use of subsumption relations, restriction classes, etc. I would submit that to a large degree, this evaluation is independent of application (although I will accept that how one chooses, for example, to do classification would depend on one’s perspective). In contrast, extrinsic evaluation of an ontology has everything to do with the driving application. Steven R. Ray, Ph.D. Distinguished Research Fellow Carnegie Mellon University NASA Research Park Building 23 (MS 23-11) P.O. Box 1
Moffett Field, CA 94305-0001 Email: steve.ray@xxxxxxxxxx Phone: (650) 587-3780 Cell: (202) 316-6481 Skype: steverayconsulting
No, it is rarely “the” application, but application is always in mind, I think, when you develop an ontology. What, after all, are requirements? Requirements for what? Represent ontology classes, relations, properties (axioms) to a certain level of granularity? What is the targeted level of granularity and why? Personally I think you always have an application (applications) in mind when you develop ontologies. Reuse of an ontology occurs when you have another application(s) in mind. The application may be very general or very specific: provide a superstructure for mid-level and domain ontologies for semantic interoperability (of systems), enable semantic search by characterizing concepts to be used for conceptual “term” expansion or content-tagging, provide a basis for enterprise engineering, provide semantic integration for these 5 databases. To me, those are applications, albeit some are generic applications. Thanks, Leo Leo, No doubt, these ontologies have specific uses and intents. And these lead to requirements (although they might be unstated). But these requirements are not dependent on the requirements of a specific application. This is why I don't agree that their development "has to fit with the development of *the* application". On Dec 18, 2012, at 2:13 PM, Obrst, Leo J. wrote: Perhaps we can weaken it a bit, to a specific use or uses? I would still say that the Gene Ontology, DOLCE, BFO, etc., have specific uses and intents, even if the use/intent is something like “act as a foundational ontology linking 3D with 4D perspectives”, or “act as an upper domain ontology for Anatomy”, or “provide an ontology of quality spaces and tropes”, etc. I would not like to quite water it down to that of any artifact. I think the following is too strong. Ontology development is always dependent on application requirements, and so development of the ontology has to fit in with the development of the application.
This is certainly true for many ontologies. But there are examples (like the Gene Ontology, Foundational Model of Anatomy) which are not being developed for a specific application. On Dec 18, 2012, at 10:36 AM, Obrst, Leo J. wrote: Ontology development is always dependent on application requirements, and so development of the ontology has to fit in with the development of the application. Across the full lifecycle. In many cases, viewed abstractly, a very successful, sound ontology can be part of a failed application. In which case, failure can taint the ontology too, or worse, the prospects and value of ontological engineering/science. That is why, in our discussions, we have talked so much about “ontology and application lifecycle”. For example, developing an ontology may require also developing or promoting a vocabulary, even multiple vocabularies, that map to the ontology. These enable user communities to use their words and phrases, their presentations, while also ensuring the representation provided by the ontology. Vocabularies can include user interfaces (forms, graphics), but also data schemas, both relational and XML-based. So it seems to many of us that ontology evaluation has to address also application use and intent. An issue that I don't see clearly in the correspondence is: How does ontology development fits into the larger life cycle of information system development? Answering requires some statement on the scope of ontologies and how they relate to other knowledge and information models. * What are the different paradigms for roles for an ontology in information or knowledge systems? As a terminology to be carried by the information model? As part of the the information model? As a means of validating the information model? Reconciling multiple information models? Other? In each case is it one model or several? If several, how are the interfaces defined? Maintained? * How does this integration into use affect the life cycle? Can we avoid too close a coupling between the ontology development and other developments so that one does not become a drag on the other. In particular how to front loading development with the work on ontology development that the applications never get built. This has been a major issue in the Health Informatics area, with enormous effort going into developing resources such as SNOMED CT and the NCI Thesaurus with much less attention to how they will be used (not to mention the related front-loaded efforts in other areas of information modelling, e.g. both HL7-v3 ). Are these issues the Summit should address? (Or have I just not looked int the right place or interpreted the comments correctly) Professor of Medical Informatics School of Computer Science TEL +44 (0) 161 275 6149/6188 <ATT00001..c>
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2013.0.2805 / Virus Database: 2637/5967 - Release Date: 12/17/12
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2013.0.2805 / Virus Database: 2637/5967 - Release Date: 12/17/12 |