Joe,
I think we're not far from agreeing. My proposal was not about endorsing one
upper-level ontology, but be ready to have different things with this status. And probably
things of dubious "ontological quality". (and yes I believe it's quite
alright, useful stuff can still be implemented on top of that).
Antoine
Subject: [Wikidata-l] 'Person' or
'human', upper ontologies and
migrating 4 million claims
Antoine
while there there are discussions in the RFC about high level ontlogies there is other
stuff happening out on the wikidata item pages.
Editors are constructing low level ontologies using ''instance of' and
'subclass of' and these are gradually creeping upwards.
'is in administrative unit' and 'located on terrain feature' are being
used to build another hierarchy of places on earth and 'part of' is being used to
build a hierarchy of places off the planet.
'occupation (person)' is becoming more important than 'instance of' in
classifying humans and 'child'
'instance of' is also being used to classify all the items derived from wikipedia
pages that don't quite fit - category pages, disambiguation pages, compound items
(describing more than one thing - like 'Bonnie and Clyde'), so tools can find
these to exclude them from queries or whatever.
Personally I can't see an awful lot of use for an upper level ontology - all the use
cases I've seen are for the lower levels. If an upper level is to be added (and
I'm sure it will - 'encyclopaedic' is close to a synonym for
'completist') then why not have all of the upper level ontologies? 'subclass
of' can be used to create a variety of upper level ontologies on top of the base
levels derived from the items we have. After all the enwp categories have three different
upper level ontologies!
Joe
user:filceolaire
On Mon, Sep 23, 2013 at 1:00 PM, <wikidata-l-request(a)lists.wikimedia.org
<mailto:wikidata-l-request@lists.wikimedia.org>> wrote:
----------------------------------------------------------------------
Message: 1
Date: Sun, 22 Sep 2013 22:24:32 +0200
From: Antoine Isaac <aisaac(a)few.vu.nl <mailto:aisaac@few.vu.nl>>
To: <wikidata-l(a)lists.wikimedia.org
<mailto:wikidata-l@lists.wikimedia.org>>
Subject: [Wikidata-l] 'Person' or 'human', upper ontologies and
migrating 4 million claims
Message-ID: <523F5200.7080704(a)few.vu.nl
<mailto:523F5200.7080704@few.vu.nl>>
Content-Type: text/plain; charset="ISO-8859-1"; format=flowed
Dear all,
First, sorry for sending an email: I want to help, but I don't have the time
required to understand how the wiki RfC mechanism work [1]. More precisely that one seems
really not the appropriate for a first dive :-(
In fact reading it I'm not even sure I understand the question anymore. To me the
original question was about the properties P31 and P279 themselves (Eric's mail still
list them as an option, albeit a popular one), ie, rather on how to represent a
classification (independent from which one is chosen). But now I see plenty hardcore
ontological discussions on the RfC page, which are indeed about getting a unified
top-level ontology...
The basic question is, can you really get a unified, perfectly structured and clean
classification of things?
I'm slightly surprised that Wikidata would go there. You want users to add
classes in the future, no? Or to use the existing wikipedia categories as a source of
classification?
In either case, you'd end up making weird inferences possible, if you apply the
formal semantics of P31 and P279 as they're defined for rdf:type and rdfs:subClassOf
[4,5]. Actually even if you invest time making a clean top-level, the lower-level parts of
the classification will probably very soon diverge from formal ontology
"meta-principles" that structure SUMO, DOLCE, BFO, etc.
And it's probably very alright, for most of your usage scenarios. Having simple,
intuitive classification semantics is possible without the full formal ontology apparatus.
Namely, you can use something that looks like rdf:type/rdfs:subClassOf, but with looser
semantics.
1. you could use something like the dc:type property from the Dublin Core framework,
instead of rdf:type. Possibly creating sub-properties of it, using a list like the one at
[7] for input.
2. You could use something like skos:broader and skos:narrower [8] for the links
between the 'looser classes'
Of course this does not correspond to formal ontological framework as in the Semantic
Web sense. But well, if the 'classification' doesn't fit a super-formal
framework, I see no reason to desperately try to shoehorn it into RDFS.
Note that I would quite disagree with the second part of the sentence from one of the
RfC-related pages [9]:
"
There is a consensus on Wikidata against creating other properties which perform this
function as it is felt a clean hierarchy of classes is in keeping with W3C recommendations
and will make it easier to use the data here.
"
First, getting a clean hierarchy won't make things easier, if you end up with a
too static/formal view on the world. Second, the feeling about the W3C recommendations is
wrong. W3C has actually pushed SKOS to allow 'softer' classifications to be
represented having to undergo the ordeals and dangers of RDFS/OWL...
But I realize all this might be regarded as questioning the decision you made earlier
on using P31 and P279 instead of the GND type, so I'm going to stop bothering you ;-)
Best,
Antoine
---
Antoine Isaac
Scientific coordinator, Europeana.eu
[1]
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
[2]
http://lists.wikimedia.org/pipermail/wikidata-l/2013-September/002815.html
[3]
http://lists.wikimedia.org/pipermail/wikidata-l/2013-September/002816.html
[4]
http://www.w3.org/TR/rdf-schema/#ch_type
[5]
http://www.w3.org/TR/rdf-schema/#ch_subclassof
[6]
http://purl.org/dc/terms/type
[7]
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
[8]
http://www.w3.org/TR/skos-primer/#secrel
[9]
https://www.wikidata.org/wiki/Help:Modeling#Hierarchy_of_classes