- From: Krzysztof Janowicz <janowicz@ucsb.edu>
- Date: Thu, 9 Jan 2014 09:11:26 -0800
- To: Raphaël Troncy <raphael.troncy@eurecom.fr>, <public-locadd@w3.org>
- CC: Pascal Hitzler <pascal.hitzler@wright.edu>
>> What does this mean exactly? You're proposing that locn should >> systematically create classes and properties in its ns and provide >> axioms (equivalentClass, equivalentProperty) with other vocab when >> necessary? More or less, yes. I think in most cases alignments are the better way to go. > Taken literally that looks like a very strong constraint. There are, of course, cases where I would not do so. E.g., if both vocabularies are from the same standardization organization, etc. > Most of us (re-)use elements from DC, DCT, SKOS, VOID, FOAF, VOAF, PROV-O, etc. FOAF is actually a good example. Many ontologies that are not even remotely about humans use FOAF these days, e.g., for building names or to assign homepages to things. From what I recall Dan broadened FOAF by removing some of the range restrictions but we are still making reasoning and integration much more difficult than it could be. >> I'm all for fighting against the obesity of having so many classes and >> properties which have exactly the same meaning, so either the resource >> is more specific and should indeed be created in the vocab or it is not >> and then, why not plain re-using the existing class/property? Well, yes but do they really have the same meaning? Most vocabularies that we see remain on the level of surface semantics. There are only a few ontologies and ontology design patterns out there that really restrict the interpretation towards the intended meaning, i.e., that really do some semantic engineering. In the most cases you just have to guess whether they actually talk about the same or not. So far, most of us have been very successful in creating a Linked Data cloud that is so convoluted that it is nearly impossible to use. On reuse/misuse of lat and long you may find this article entertaining: http://stko.geog.ucsb.edu/location_linked_data . For instance, it uses Linked Data to show that humans never landed on the Moon. IMHO, you can just define whatever you want in your local namespace, give it as much semantics as possible (or necessary) and then have an additional layer of axioms that establishes alignments. Many companies will (or do) it this way anyway because they are afraid of giving up control. Will there be a FOAF in 5 years from now? Best, Krzysztof On 01/09/2014 03:35 AM, Raphaël Troncy wrote: > Dear Krzysztof, > > Le 09/01/2014 07:03, Krzysztof Janowicz a écrit : >>>> About how to model GNIS IDs and other geographical identifiers, as I >>>> said in [1], a possible option would be to re-use adms:Identifier [2]. >> >> IMHO, other vocabularies/ontologies should be aligned but not reused in >> the sense of a fixed integration. > > What does this mean exactly? You're proposing that locn should > systematically create classes and properties in its ns and provide > axioms (equivalentClass, equivalentProperty) with other vocab when > necessary? > > I'm all for fighting against the obesity of having so many classes and > properties which have exactly the same meaning, so either the resource > is more specific and should indeed be created in the vocab or it is not > and then, why not plain re-using the existing class/property? > > Raphaël > -- Krzysztof Janowicz Geography Department, University of California, Santa Barbara 5806 Ellison Hall, Santa Barbara, CA 93106-4060 Email: jano@geog.ucsb.edu Webpage: http://geog.ucsb.edu/~jano/ Semantic Web Journal: http://www.semantic-web-journal.net
Received on Thursday, 9 January 2014 17:12:06 UTC