W3C home > Mailing lists > Public > public-vocabs@w3.org > May 2014

Re: Are there any plans to develop an OWL version of QUDT?

From: <martin.hepp@ebusiness-unibw.org>
Date: Thu, 8 May 2014 09:35:22 +0200
Cc: Holger Knublauch <holger@topquadrant.com>, W3C Web Schemas Task Force <public-vocabs@w3.org>, Ralph Hodgson <rhodgson@topquadrant.com>
Message-Id: <E1E4AFE9-CC50-417E-817A-0B10DE79BB3F@ebusiness-unibw.org>
To: Simon Spero <sesuncedu@gmail.com>
Let's put it short:

If a vocabulary is outside OWL DL just because the model is sloppy or the modeler did not understand the constraints of OWL DL, it is typically desirable to aim at staying in OWL DL. The advantage is that the vocabulary can be used in OWL DL environments without turning an otherwise OWL DL knowledge-base into OWL Full just by adding the vocabulary. That was, btw, the motivation to make extra effort for keeping GoodRelations in OWL DL - so that people who use it in more controlled settings (e.g. internal product data management) can use an OWL DL environment.

If, however, OWL DL constraints make modeling the domain cumbersome or require the removal of options, then being in OWL Full is IMO better than crippling the vocabulary to OWL DL.

For instance, the frequent pattern in schema.org to allow both typed entities and text values for certain properties is IMO a feature, but can be handled in OWL DL only by cumbersome solutions. One could use a owl:DatatypeProperty with the range of the union of xsd:anyURI and xsd:String (from the top of my head I think that even that is outside of OWL DL, but I am not sure).

Or take sub-properties of owl:DatatypeProperties. I may be wrong here again, but afaik this has been removed from OWL 2 DL, while it was in OWL 1 DL and is a heavily used pattern in GoodRelations. For instance, you may want to make foo:hasXYZ code a subproperty of foo:hasCode so that you can easily abstract from the exact code type in queries.

The datatype xsd:time is no longer in OWL 2 DL, and while I understand the logical argument for removing it, it is a feature we use heavily for modeling opening hours.

I personally think that the step from OWL 1 to OWL 2 was not done well, since weakening an already weak user base by introducing constraints that popular adopters cannot meet is problematic.

Ontologies are interfaces between human minds, human minds and computer systems, and computer systems and computer systems. When making design choices, we have to look at existing data structures, cognitive skills of humans, and computational characteristics of the representation all at the same time.

OWL 2 DL is better for computers. It is arguably not better for Web ontologies and the Semantic Web ecosystem.

Martin

-------------------------------------------------------
martin hepp
e-business & web science research group
universitaet der bundeswehr muenchen

e-mail:  martin.hepp@unibw.de
phone:   +49-(0)89-6004-4217
fax:     +49-(0)89-6004-4620
www:     http://www.unibw.de/ebusiness/ (group)
         http://www.heppnetz.de/ (personal)
skype:   mfhepp 
twitter: mfhepp

Check out GoodRelations for E-Commerce on the Web of Linked Data!
=================================================================
* Project Main Page: http://purl.org/goodrelations/




On 08 May 2014, at 03:41, Simon Spero <sesuncedu@gmail.com> wrote:

> On May 6, 2014, at 8:19 PM, Holger Knublauch <holger@topquadrant.com> wrote:
> 
>> +1
>> 
>> OWL DL is IMHO an academic niche language while most real-world projects that I have seen don't need or want its limitations. I also find it alarming if people argue that we should not use RDF features only because their favorite ontology editor and API cannot handle RDF. Being limited to OWL DL was a design decision that those tools made, while other tools have no problems with RDF.
>> 
>> Holger
> 
> You raise some good points;  as I have often stated (pace pfps) , I consider Description Logics to be rather weak tea, and whilst they are very well suited to a certain class of problem, I greatly prefer non-montonic, higher order systems for general KR problems. 
> 
> However, some of the points you make require some caveats.   
> 
> Firstly:  
>    If a problem falls within the class of problems that OWL 2 DL can handle, it is good to stay within the constraints of DL.  Going to OWL Full should be made on a carefully considered basis of costs vs. benefits.  
> 
> Secondly: 
>       The mapping from OWL to RDF is explicitly only defined for OWL-DL.  It is not possible always possible to parse an OWL-2 ontology that does not follow some of the DL restrictions.  One can make a best effort, but the mapping from OWL 2 Full to RDF is a one way trip, as the OWL to RDF mapping specifications makes clear in the first paragraph of the first section. 
> 
> Thirdly:  
>         The vocabularies being discussed were inconsistent at every level of the RDF, RDFS, and OWL chain, and contained basic type errors. I do not know what tools were used to create them, but whatever tool it was could not do basic consistency checking.   
> 
> Fourthly:  
>      If a vocabulary or tool claims to comply with  standards/(W3C recommendations), it is very important that the vocabulary/tool should comply with those standards. If a vocabulary tool is intended to follow a standard, but is in fact non compliant, then it is a major bug.  
> 
> As Ralph said, he considers the problems I reported to be potentially serious issues, and will be looking in to them in detail.
> 
> Simon
Received on Thursday, 8 May 2014 07:35:50 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:29:41 UTC