Re: Reflexivity and antisymmetry uses cases?

On Jan 17, 2007, at 2:52 PM, Holger Knublauch wrote:

>
> Christine Golbreich wrote:
>> Hi Holger
>>> extra features may be more driven by theoretical advances in DL
>>> reasoners than by real user requirements
>> 1)You may find several of these requirements in different papers  
>> that I
>> presented at various workshops/conf since 2003 (sorry to advertise)
>
> [snip]
>
> Thanks, so you seem to be number 4 on the list of people who are  
> interested in these new features.

I think Christine and Alan can be taken as representative of large  
sections of the bio-ontology community.

> I guess each of the features currently discussed for OWL 1.1 has  
> received some support by someone. For example, I am sure some  
> people would like to see support for probabilistic information,  
> better reification support, or if OWL could do mathematical  
> calculations, up to automatic unit conversions.  There will be some  
> use cases for all this (and actually I have heard the example  
> features above many times, yet they likely won't make it into OWL  
> 1.1).
>
> The main point I am raising here is: how do we limit what feature  
> should be included and what should rather not be.  Does it depend  
> on the number of potential users of the feature, or whether  
> something is easy to implement, or both?  I am worried that 1.1 is  
> already adding too much, alienating the capabilities of OWL further  
> and further from average users.

I've spoken to quite a few alienated users, and non-users who didn't  
get past the initial alienation stage. I don't think adding new  
incremental features such as anti-symmetry will add to existing  
alienation - these aspects can simply be ignored. I think people may  
be alienated by a number of things, all of which have been duly noted  
already - these are all-or-nothing kind of things, rather than a case  
of feature creep. From a non-ontology versed bioinformatics  
programmer point of view, dealing with an OWL document involves  
figuring out the whole OWL-RDF-XML stack, or one of the other myriad  
serialisations. Also, to someone coming from this point of view, it  
takes a while to explain why it's so involved to state something  
simple like X part_of Y.

Don't want to sound like I'm complaining - this has all been said  
before - and OWL seems to be making tremendous advances despite these  
difficulties - you just seem to be asking for anecdotal evidence,  
rather than documented requirements such as those Christine provided.  
 From many people's POV the main hurdle is using OWL at all, the  
addition of backwards compatible features won't be much of an extra  
turn off.

If you want my vote, I'd like to echo Christine's and Alan's  
endorsement of transitive_over/property chains. I can provide  
numerous examples of where these would be required. I believe Alan  
Rector has long been in favour of these - they have been in Grail/ 
GALEN for some time, I think. He could perhaps provide some examples  
too.

>   On the other hand, it is clear that features like user-defined  
> datatypes would make OWL more attractive to user communities that  
> currently cannot work with OWL.  Perhaps it would be useful if the  
> working group would come up with an informal Use Cases document  
> that illustrates why certain features have been requested (maybe  
> such a list already exists somewhere?).  Other working groups such  
> as RIF even take a use cases list as their first deliverable.
>
>
>> Are you sure for example that in the ontology developped by  
>> Olivier et al..
>> in Virtual Soldier (the Protégé - DARPA project ) they did not  
>> have such
>> things ?
>
> This project was one among hundreds of projects that used Protege.   
> Not every OWL user is creating medical domain ontologies.  I don't  
> remember a lot of requests for something like owl:SelfRestriction  
> on our mailing lists.  Maybe I am wrong.  If I see many more  
> compelling examples, I am easily converted into a supporter of  
> reflexivity and antisymmetry (for me they are actually trivial to  
> implement - this is not the issue here).
>
> BTW: The classes owl:IrreflexiveProperty, owl:AntisymmetricProperty  
> could easily be defined in another namespace outside of OWL 1.1 -  
> if a certain community needs these features then they can define  
> and import their own ontology with these extras.  Ontology editors  
> would handle these types just like any other rdf:Property  
> metaclass.  Yet the new types don't need to bloat the OWL spec and  
> education documents.  Just a thought... feel free to crucify me.
>
> Holger
>
>

Received on Thursday, 18 January 2007 00:12:19 UTC