Re: follow ups on EIIF

Some feedback from Vocamp, notes from me  interspersed with comments from
Norman
(Norm, please note the one line of comment in CAPS about the namespace)
cheers
PDM


On Wed, Sep 16, 2009 at 10:00 PM, Norman Gray <norman@astro.gla.ac.uk>wrote:

>
>
> On 2009 Sep 16, at 20:31, Paola Di Maio wrote:
>
>  1. according to some experienced participants  (but not domain experts)
>> who
>> are giving feedback on the diagram, the process/method
>> of formulating the diagram first and deriving a potential vocabualry/owl
>> file from it is good, the way to go
>>
>
> Yes.  Whenever folk talk about developing ontologies they always emphasise
> the importance of this initial capturing of a model from domain experts,
> through discussion.  That stage can be done without once mentioning the
> O-word.
>
> Turning that diagram into an ontology or vocabulary is potentially the hard
> bit.
>
> If what's required is a controlled vocabulary -- that is, basically a tree
> or trees -- then that requires little more than expressing the tree
> structure using SKOS.
>
> The EIIF model that you showed does look more like an ontology (especially
> given the typed properties we were talking about).  That can be turned into
> a lightweight ontology, perhaps using little more than RDFS, or into a much
> more heavily structured thing.  Which of those you do depends very heavily
> on what the ontology is to be used for, and it's important to be clear about
> that up-front. An Ontology is not an end in itself: there are some questions
> that you want answered, and some conclusions that you want to be able to
> draw.  As Ed illustrated in his talk, it's important to have some test
> relations identified ahead of time.
>
>  2. the Cabinet in the UK is a bit stuck in RDFizing their vocabulary,
>> posted
>> recently to ths list, cause they havent decided on the namespaces yet,
>> among
>> other issues,  Peter who attended the vocamp and works for the government
>> is
>> going to try to solicit them gently, he agrees that some things cannot
>> wait
>> for govenrments to get unstuck . If anyone decides to rdfize vocabularies
>> with temporary namespaces could help move things along. Let's see if
>> something happens
>>
>
> I agree that it's important to get things under way, and that a temporary
> namespace might help get things moving until the vocabulary gets some
> traction.  But beware: the evidence that the vocabulary has some traction
> may be that it's being used sufficiently widelyl that you can no longer
> change the namespace.  This happened with FOAF and with SKOS.  The FOAF
> namespace is http://xmlns.com/foaf/0.1/ and will remain so no matter what
> version FOAF has -- they can't change the version number.  A similar thing
> happened with SKOS.
>
> I'm not saying there's an easy answer to this, just that no-one should
> blithely say 'och, we can change the namespace later...'.
>

OKAY, BUT WE CAN START WORKING AND THEN SLIP IN THE NAMESPACES JUST IN TIME
BEFORE PUBLISHING THE  VOCABS PERHAPS PROVIDED THEY ARE READY, AND IF NOT,
WORRY ABOUT IT LATER?

>
>  3. Either way, to map the vocabulary derived from our model to
>> vocabularies
>> from other efforts would be the right thing to do,  however mapping
>> techniques for vocabularies are still not comprehensive established tested
>>
>
> Yes, absolutely.
>
>  4.  To map two vocabularies, say the EIIF to the one produced by the
>> Cabinet
>> Office, would be straighforward, but to map multiple vocabularies, when
>> each
>> term has multiple kind of relations to other terms, is likely to yield an
>> unmanageable #'combinatorial explosion' (as put by Gordon Dunsire), the
>> possible way out is the Hub and Spoke approach, whereby one vocabulary
>>  (in
>> our case the  EIIF model vocabulary) is consdered the central reference
>> and
>> each vocabulary only maps to the central vocabulary, could be a solution ,
>> there are shortcomings there as well, ...
>>
>
> I don't think I'd go as far as 'straightforward' to map two vocabularies,
> but it's certainly easier than doing several.  As Gordon said, the
> hub-and-spoke model is an effective way to proceed; I was concerned to
> stress that it doesn't rule out adding some pair-wise mappings as well.
>  Jenny Ure can supply further models. The key thing here, I think, is that
> this is the web -- there are many ways it can work.
>
> I hope these extra remarks help.
>
> All the best,
>
> Norman
>
>
> --
> Norman Gray  :  http://nxg.me.uk
> Dept Physics and Astronomy, University of Leicester, UK
>
>


-- 
Paola Di Maio
**************************************************
Networked Research Lab, UK

***************************************************