Re: ISSUE-160: Allowing collections in semantic relationships

Hi Tamara!

I am not meaning "machine" indexing as I wrote that, IMHO, SKOS is used 
as a basis to link humans (building/using their conceptual models and 
intuition) to machines (building/using their data models).

SKOS is not right for machine indexing for different reasons:

1) Thesaural hierarchies may seem to be logical hierarchies (partitions, 
dependencies, etc.) BUT they are often (not always, I agree) based on 
the rule "the (human) user will gratefuly receive everything indexed 
with all entries below a given entry in the hierarchy". This document 
better explain this:
http://www.nlm.nih.gov/mesh/meshrels.html

2) Thesaurus terms can be "precise" concepts but they can be 
pre-coordinated (or even "blurred") concepts (descriptors) used for 
classification purposes. No automated "reasoning" is possible on this 
kind of entries.

3) SKOS lacks (and it is VERY far from) the Natural Language parsing 
rules needed to automatically parse, interpret, abstract, index a document.

For me, SKOS is very important because it could encompass all the 
necessary data to make User Interfaces efficient for indexing/retrieval 
systems (and may be other kind of systems but indexing/retrieval is 
already a very tough mission). Example of questions were SKOS should help:
* How do you group documents in meaningful way for users?
* How do you sort them to present the most useful first?
* How do you help users specifying/discovering their needs in a dynamic 
dialogue with the available information?
* How do you make search on precise topics?
* How do you provide "horizontal searches" for linked information?
* ...

So "distributed indexing and retrieval" by humans using machines is my 
interest in SKOS: I would understand that other have other interests (at 
least, I try to be clear why I am here!). SKOS is for the information 
exchanges within this "whole system": applications providing benefits of 
SKOS are to be defined and built... I would like to participate to this.

I realy agree with the fact that all this should remain as simple as 
possible and built over real life experience: current standard, as is, 
is complete enough to start implementations (I am doing this daily). 
Implementations will evolve with needs and the standard will follow!

I also understand (correctly?) your message Tamara about the fact that 
thesauri are built for some purposes (like computer tools and even 
standards like SKOS!). And that it is extremely difficult to "share" 
thesauri because slightly different purposes create very different 
thesauri. Furthermore, that thesaurus makers rarely precisely explain:
* the needs of their users,
* the internal mechanisms of their information system,
* the rules they follow for their decisions.
You propose that some of this should appear in the standard to support 
"distribution" of thesauri ? I would agree if this enables automated 
treatments to better use a given thesaurus in different environments. 
Big challenge, may be not immediatly adressable. OWL could be a good 
tool to model thesauri management attributes (not their content, their 
characteristics) and classify them...

For me, without common purposes in a distributed system, it is very 
difficult to have a common KOS: well defining the purposes is the basis 
to specify the KOS.

I hope we can share and build a momentum: please find below a conference 
where it could be fine for the SKOS enthusiasts to meet (any other 
sooner???).

Have a nice day!

Christophe Dupriez

==================== 2nd CALL FOR PAPERS=======================
International UDC Seminar 2009

CLASSIFICATION AT A CROSSROADS - MULTIPLE DIRECTIONS TO USABILITY
The Hague, 29-30 October 2009

The 2009 Seminar will explore how new developments in information 
standards and
technology influence and affect applications and services using 
classification
and its relationship to other systems.

Two eminent keynote speakers will address the conference: Prof. Karen 
Markey,
School of Information, University of Michigan, a well known author and
researcher in the area of improving online information systems and Dan 
Brickley,
an advocate and developer of W3C Semantic Web technologies, who founded and
chaired the Semantic Web Development Group from 1999-2005.

The conference programme will highlight many ways in which 
classification use
can be improved. Attention will be paid to the applications of 
classification
schemes, UDC in particular, in supporting multilingual access, 
user-friendly
representations of classification in resource discovery and application 
across
distributed systems.

Papers are now invited from classification researchers and practitioners 
on the
following topics:

1. classification and semantic technologies
2. classification in supporting information integration
3. verbal and multilingual access to classification
4. classification authority control and library systems
5. visual representations/interface to classification
6. experiences with classification outside the traditional library 
environment

Contributions may include conference papers and posters. To read more 
about the
conference and its theme and to submit your abstract, please go to the
conference website: http://www.udcc.org/seminar2009/index.htm.

Contributions will be reviewed by the Programme Committee consisting of an
international panel of classification experts. Proceedings of the 
conference
will be published in Extensions and Corrections to the UDC, 31 (2009). A
selection of the best papers will be considered for publication in 
Knowledge
Organization, the leading peer-reviewed journal in the field.

IMPORTANT DATES:
01 Feb 2009    Abstract submission deadline
01 Apr 2009    Notification of acceptance
31 Jul 2009    Full papers/posters submission

CONTACT:    seminar2009@udcc.org

ORGANIZER: The seminar is organized by the UDC Consortium (UDCC) and 
hosted by
The National library of The Netherlands. UDCC is a non-profit organization,
based in The Hague, established to maintain and distribute the Universal 
Decimal
Classification (UDC) and supports its use and development.




Stephen Bounds a écrit :
>
> Hi Tamara,
>
> This is almost exactly my interpretation of SKOS as well.
>
> I also support Antoine's position of keeping things as simple as 
> possible, with the possibility of well-defined extensions.
>
> Cheers,
>
> -- Stephen.
>
>
> Lopez, Tamara wrote:
>> Hello all,
>>
>> Sorry to jump in here late, but I've only just had a moment to
>> properly work through this thread, and I can't resist throwing a
>> couple of observations on the pile...
>>
>> Since I've been experimenting with SKOS, I've wondered just *what* it
>> intends to exchange.  Does it exchange a 'tool' for organising
>> resources or 'data' about such a tool?  My impression, which has
>> largely been confirmed by this thread, is that it is the latter.
>> If so, then it seems to me that the use of semantic terms like
>> collection, broader than, narrower than *are* primarily to be used to
>> articulate documentary (and therefore presentational) information
>> about KO systems, and that labels for these concepts may have only a
>> tenuous link to what these concepts mean or how they behave in the
>> actual KOS itself.
>>
>> Is my understanding correct?
>>
>> If so, my second question is , what uses should these SKOS
>> representations be expected to support? We've seen some examples of
>> linking to create a network of sorts between similiarly
>> pre-synthesized concepts in other schemes (I'm primarily referring to
>> the gene database). These networks are formed between the URIs, and
>> do not seem to be doing very much with the relationships expressed in
>> the SKOS (if my understanding is correct).  What is the depth of such
>> a network, and doesn't it depend quite heavily on tacit agreement
>> between different organisations' interpretation of 'Concept'?
>>
>> Can a SKOS file be used  as Christophe Dupriez suggested for
>> distributed (I assume machine) indexing and retrieval?  My impression
>> is no.  In my own use of the standard for a simple indexing/retrieval
>> system[1], it was the unformalised principles of how the tool worked
>> that needed to be exchanged (even between areas of our application
>> stack) in order to digitally recreate the KOS.
>>
>> In our implementation, a lot of the 'knowledge' remained in our
>> heads, was tacitly expressed via material aspects of the format[2] or
>> was written into surrounding software.  In fact, as Aida has
>> suggested, we primarily used our SKOS as an artifact of modeling - a
>> diagram of sorts that people could quickly grasp and talk (or write)
>> around while implementing algorithms that employed various principles
>> of the taxonomy.  In fact it worked well in this role, but I didn't
>> finish the project feeling that the SKOS we had produced was a tool,
>> nor that it alone could be shared in a meaningful way with others who
>> wanted to use it for a similar purpose[3].
>>
>> If my observer take on the situation is correct, then I'd just like
>> to conclude by agreeing with the sentiments expressed in this thread
>> about the need to look at older forms and systems for some guidance
>> as SKOS moves forward and to suggest that the primer should carefully
>> communicate  the utility of the standard to potential adopters.
>>
>> Regards, Tamara
>>
>> [1] built using an earlier version of the Primer and Recommendation
>> and expressed using SKOS RDF/XML.  Indexers were applying
>> pre-coordinated subject strings via notation to resources, strings
>> were composed of three terms, second and third level terms could
>> belong to more than one 'class', and in retrieval, the class order
>> was not fixed - users could access the hierarchy from any point. [2]
>> the presence of children or parents, idiosyncratic use of attributes,
>> and even the direction in which we designed the SKOS  tree to be
>> processed. [3] And this was partly due, I'm sure, to weaknesses in
>> our interpretation of the standard- not to weaknesses in the standard
>> itself.
>>
>> -- Tamara Lopez Centre for Computing in the Humanities King's College
>> London 26-29 Drury Lane London WC2B 5RL (UK) Tel: +44 (0)20 78481237 
>> http://www.kcl.ac.uk/cch
>>
>>
>>
>>
>
>
>