On 7 April 2011 19:32, Egon Willighagen <egon.willighagen at gmail.com> wrote:

> On Thu, Apr 7, 2011 at 2:52 PM, Christoph Helma <helma at in-silico.ch>
> wrote:
> > That looks very document centric, not sure if all of these terms are
> > applicable e.g. to models and algorithm.
>
>
This was also my first impression.


> I guess ontology stuff can get a bit philosophical... if you consider
> the specification of a particular model or algorithm a 'document',
> then it sounds suitable for me. In fact, this probably is the case,
> not? The service itself can be down for any particular reason, even if
> it is published or retracted.
>
> > Would it be possible to allow
> > only a subset of these terms in our metadata?
>
> Sure.
>

I don't have preference towards any of the listed ontologies, of course no
need to enforce all defined properties.


>
> Nina, does the OpenTox ontology distinguish currently between a model
> / algorithm instance and its specification?
>

Yes, algorithms and models are different resources by design.  Models are
created by applying an algorithm to particular data (e.g. linear regression
equation is a model, derived by application of a linear regression
estimation algorithm -  could be Ordinary Least Squares, but also any
other).

Thus, the algorithm could be considered a specification of the model, but
also a processing resource itself.  Specification of the algorithm itself is
less strict, details are expected to be included in the algorithm RDF
representation  (similar to your suggestion to use bo:instanceOf  for
descriptor calculation, relying on BO ontology for algorithm description).
Generic algorithm types (e.g. regression, classification, etc.) are also
defined in the algorithm types ontology and linked from there.


>
> (This reminds me about the discussion of http://example.org/SomeTopic
> is the URI of a webpage or the URI of the resource described on that
> webpage :)
>
>
Arrrgh ... we'd better take more pragmatic route in OT, we are dealing with
live online resources, addressable by URIs. They are both informational
(description of the resource) and non-informational (e.g. the model resource
itself is richer than its RDF description), to use the terminology from [1].

HTTP URLs were initially designed to identify networked resources, not
abstract constructs, so the semantic community is kind of abusing URIs,
however (sometimes) taking an advantage of the networking infrastructure
that allows to resolve them.

Otherwise ... "the problems of naming and addressing have a tendency to get
philosophical" [2] (even in technology).  Naming and locating things are
different concepts.  Yet more different is relating a name to its meaning
(as in the above http://example.org/SomeTopic - does the meaning equate to a
description of the resource or to the resource itself ;)  But this is too
much of off-topic already :)

Nina

[1] http://www.w3.org/2001/tag/doc/httpRange-14/2007-05-31/HttpRange-14
[2] J.Day, Patterns in network architecture, Chapter 5. Background on Naming
and Addressing.

Egon
>
> --
> Dr E.L. Willighagen
> Postdoctoral Researcher
> Institutet för miljömedicin
> Karolinska Institutet (http://ki.se/imm)
> Homepage: http://egonw.github.com/
> LinkedIn: http://se.linkedin.com/in/egonw
> Blog: http://chem-bla-ics.blogspot.com/
> PubList: http://www.citeulike.org/user/egonw/tag/papers
>