W3C home > Mailing lists > Public > public-esw-thes@w3.org > July 2006

Re: questions about notations / sizes / coordination

From: Jakob Voss <jakob.voss@gbv.de>
Date: Thu, 20 Jul 2006 12:11:47 +0200
Message-ID: <44BF56E3.2000204@gbv.de>
To: public-esw-thes@w3.org

Hi Alistair,

You wrote:
> Somebody is awake and we are definitely interested :)

Great. I thought everone is on summer-holidays :-)


> Regarding notations, I have suggested a "skos:notation" property, whose
> range would be typed literals [1]. I suggest typed (rather than just
> plain) literals because the datatype then defines the context in which
> the notation is uniquely identifying, i.e. you wouldn't get notation
> collisions. We ought to have an item on the SKOS proposals and issues
> page for notations - I'll do that asap.

I don't mind if it's

<skos:prefLabel xml:lang="art">...</skos>

Why not this way? You don't have to change anything.

<skos:notation>...</skos:notation>

Your suggestion (by the way then xml:lang is meaningless because
notation are non linguistic and you have to introduce a new requirement
about uniqueness of notations), or

<dc:identifier>...</dc:identifier>

But I'd like to see a decision soon.


> Regarding the "size of a collection", you could first use the rule [(?x
> skos:subject ?y) (?y skos:broader ?z) -> (?x skos:subject ?z)] to
> construct a new graph, and then evaluate the following SPARQL query
> against the new graph:
> 
> SELECT ?x ?y WHERE { ?x skos:subject ?y }
> 
> For each distinct binding of ?y (i.e. for each concept), the number of
> result bindings in which that value occurs will give you the "size of
> the collection" for that concept.

Wait! I understand what you are saying and your right, but I am trying
to build real world digital library systems with hundreds of thousands
of records so I cannot use SPARQL and other semantic-web-toys. There are
several collections in libraries that may share a Concept Scheme but you
cannot easily dig into their system. All you get is the number of
records that are indexed with a concept and an URL to query the OPAC
with this concept. It's not the question how to determine the numbers
but how to encode the results.

Here is an updated sketch:

<skos:recordCount>
  <skos:documentCollection
        rdf:resource="http://www.oclc.org/worldcat/" />
  <skos:usedConcept rdf:resource="#ddc33" />
  <skos:recordNumber>10298</skos:recordCount>
  <skos:expandedRelation
        rdf:resource="http://www.w3.org/2004/02/skos/core#narrower" />
  <dc:date>2006-07-20</dc:date>
</skos:recordCount>

This would say that in OCLC WorldCat (http://www.oclc.org/worldcat/) at
2006-07-20 there were 10298 records with skos:subject #ddc33 or narrower
concepts. Another real world example:

<skos:recordCount>
  <skos:documentCollection
        rdf:resource="http://www.flickr.com" />
  <skos:usedConcept
        rdf:resource="http://www.flickr.com/photos/tags/birthday/" />
  <skos:recordNumber>767884 </skos:recordCount>
  <dc:date>2006-07-20</dc:date>
</skos:recordCount>


> Regarding the issue of "combinations" of concepts, there is an issue
> raised specifically for this problem [4]. I have written some rough
> notes at [2] and [3], that's as far as I've got :) There are some subtle
> issues involving when the order of coordination/combination is/is not
> significant, and how coordination/combination interacts with subject
> indexing. But supporting coordination/combination is definitely in scope
> for SKOS as far as I'm concerned.

Thanks for the links. well, coordination is not trivial :-(

Greetings,
Jakob
Received on Thursday, 20 July 2006 10:12:10 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:54 GMT