[SKOS] Re: Comments on "SKOS: requirements for standardization"

Hi Bernard,

> Antoine
>
> see comments in line.
>
> Antoine Isaac  wrote on 19/11/06 21:10:
>
>> Hi Bernard,
>>
>>>
>>>
>>> some comments on
>>> http://isegserv.itd.rl.ac.uk/public/skos/press/dc2006/camera-ready-paper.pdf 
>>>
>>>
>>>
>>> On the whole I agree with much of this, though I'm concerned (and 
>>> probably a little confused) about the use of the term 'index' as 
>>> well as some of the underlying assumptions of SKOS.
>>
>>
>>
>> This use seems to me quite common in the information science world. 
>> What makes you really unconfortable with?
>
>
> I'm not from that world so the term is unfamiliar to me. I don't think 
> we can assume that all our readers will be from that world so I think 
> we should spell out what our WG (and information science) mean by an 
> 'index'.

All right, I'll try to be cautious with that!

>
> <snip>
>
>>> '3.  Anticipated Software Architecture'
>>> This section seems to omit what seems to me will be a requirement: 
>>> the transformation of an existing taxonomy/ontology into a SKOS 
>>> taxonomy. The transformation may be batch or dynamic, but I believe 
>>> will be necessary.
>>>
>> I think you're right about the need for a conversion. However, I'm 
>> not sure this has repercussions on SKOS requirements. Could you be 
>> more precise about which kind of req you envisioned ?
>
>
> OK, let me give you a couple of concrete examples:
> 1) The Gene Ontology. There's a version of the Gene Ontology expressed 
> in OWL. I'd like a version in SKOS. I want a way of 
> transforming/transcoding GO-OWL so that I can use it in both of my use 
> cases. I have no control over GO-OWL, and have no idea how often it 
> changes, so I'd like the transcoding to be dynamic, such that when I 
> ask a service for a SKOS version of GO, it transcodes the current 
> version into SKOS for me.
>
> 2) The Sun Unified Product Taxonomy (UPT). The UPT is expressed in 
> RDFS, and I'd like a version in SKOS (for use in both use cases). 
> However, Sun has control over the UPT and can produce a parallel 
> version in SKOS whenever the RDFS version changes. So, in this case 
> all that's required is a batch conversion.

OK, this is very clear to me. I however wonder wether this will 
translate into specific requirements for SKOS. For me that would rather 
concern the features of the producing SKOS (unless you say that these 
services would ideally be relying on technologies such as XML and RDF, 
which gives a motivation for SKOS being encoded in such languages)

Anyway, the situation you mention is very interesting to mention in your 
use case I think. Which makes me wonder wether you would find some 
location in our format to put it in. Do you think point 1.4 in the 
Application would be fitting, at least for the (more interesting) case 1?

Cheers,

Antoine

Received on Tuesday, 28 November 2006 10:13:59 UTC