W3C home > Mailing lists > Public > semantic-web@w3.org > May 2013

Re: SKOS question: car models and make

From: Andrea Splendiani <andrea.splendiani@iscb.org>
Date: Tue, 7 May 2013 12:48:02 +0100
Cc: "semantic-web@w3.org" <semantic-web@w3.org>
Message-Id: <104C371B-92D1-4535-B290-828338E868F0@iscb.org>
To: Jeremy J Carroll <jjc@syapse.com>

Not an expert of SKOS either.
I think you will need an ex-SKOS property to characterize the link between genes and alleles. So having two schemes + property seems a good choice.
I would consider collections rather than schemes. At least for genes, you have a flat space characterized by belonging to the same organism. But this may be different in your specific use case.
One thing I would pay some attention at is your definition of alleles and genes.
A gene is a bit semantically polymorphic. You can see it as a unit of genetic information which you may inherit (pretty much in line with the conceptualization behind alleles), but you can also see it as a nucleotide sequence. Then imagine you talk about meta-genomics, alleles are not strictly related to genes anymore. 
By the way, I'm not even sure that the term allele makes sense for simpler organisms (say, a virus).

If you are dealing with humans, alleles and genes are more coherent, but still depending on how you see the gene, the relation between one and the other may vary.

Getting back to Steven, rather than considering it a meronymy relation, I would consider a functional interpretation in terms of parameters/values. So an allele would be linked to a gene by something sounding like "hasValue".
In any case, all these properties are out of SKOS. If you strictly need SKOS, than probably narrower is the only possible relation (with some great level of information loss and distortion).


Il giorno 06/mag/2013, alle ore 22:54, Jeremy J Carroll <jjc@syapse.com> ha scritto:

> I am trying to model some vocabularies to do with genes and alleles in SKOS.
> For the sake of a wider audience, I can reformulate the question in terms of cars.
> If I go to, e.g.
> http://www.fueleconomy.gov/feg/findacar.shtml
> and enter a year in the form (e.g. 2008)
> then I get a menu which is a choice from a vocabulary of makes, and I can make a selection (e.g. Volkswagen or Honda)
> then I can choose a model (e.g. New Beetle Convertible or Accord )
> Notice that it is an error if the UI presents Accord as an option for Volkswagen
> How should I model this with SKOS?
> With a naive mapping to SKOS from what we currently have, I would get different concept schemes for makes (genes) as opposed to models (alleles), and a property on each model (allele) showing the related (gene).
> What is the 'right way' to do this?
> Jeremy J Carroll
> Principal Architect
> Syapse, Inc.

Received on Tuesday, 7 May 2013 11:48:29 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:45:33 UTC