Fwd: "zero relations" in dataset mappings (has no matching entity in)

Forwarding this interesting discussion to the SKOS list...
Starts at http://lists.w3.org/Archives/Public/public-xg-lld/2011Jan/0113.html

Antoine

-------- Original Message --------
Subject: 	AW: WG: "zero relations" in dataset mappings (has no matching entity in)
Date: 	Fri, 28 Jan 2011 10:44:30 +0100
From: 	Neubert Joachim <J.Neubert@zbw.eu>
To: 	Bernard Vatant <bernard.vatant@mondeca.com>, Emmanuelle Bermes <manue.fig@gmail.com>
CC: 	Antoine Isaac <aisaac@few.vu.nl>, public-xg-lld <public-xg-lld@w3.org>



Hello Emanuelle, hello Bernard,
Thanks for the clarifications about the different cases, Bernard - my initial post subsumed "false positives" under #2, whereas it is better and more exactly treated as the case Emanuelle deals with (#1).
I think Bernards proposition would nicely cover both cases, and in my eyes it could fit well with the SKOS matching properties.
However, when vocabulary V2 in the statement
X skos-plus:noMatchIn V2
is enhanced, the statement may turn wrong.
(Maybe we here have a general mismatch of the Open World Assumption in the Semantic Web and the closed world of authorities in the library world.)
Anyway, to deal with this - simply reporting the fact, no reasoning intended ... -, it could be helpful to minute the time when the statement was made, or the version of V2, if V2 is versioned in some way.
Any ideas about this?
About the generalization of SKOS mapping properties, in order to avoid false owl:sameAs: The latter is clearly a requirement, which we discussed in the Authority Cluster too. There we came up with the proposition to use umbel:isLike. I'm not sure if SKOS (plus) should extend the area where it deals with owl:Things in general - even though it's done a very good job with its labeling und annotation properties. I wonder what other people think about this.
Cheers, Joachim

     -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-----
     *Von:* Bernard Vatant [mailto:bernard.vatant@mondeca.com]
     *Gesendet:* Donnerstag, 27. Januar 2011 20:15
     *An:* Emmanuelle Bermes
     *Cc:* Neubert Joachim; Antoine Isaac; public-xg-lld
     *Betreff:* Re: WG: "zero relations" in dataset mappings (has no matching entity in)

     hello all

     two points :

     First the case made by Emmanuelle (#1) is not the same as the one made by Joachim (#2)

     #1 "Y (in vocabulary V2) is not exactMatch of X (in vocabulary V1)"
     #2 "X (in vocabulary V1) has no match in vocabulary V2"

     #1 is an assertion of exclusion of a false positive (e.g., false assumptions based on homographs),
     whereas #2 is the assertion of a global absence of match.

     The logical relationship between the two is, for a given X

     #2 <=> forAll Y, #1

     Second, using X owl:differentFrom Y to express #1 is a bit misleading.
     It means that X is not the same as Y (in the very strong sense of owl:sameAs)
     That does not mean they could not be matched.

     (X skos:exactMatch Y) and (X owl:differentFrom Y) are not necessarily inconsistent triples.
     In fact I would say that most of the times, concepts matched, even by skos:exactMatch are not the same is the sense of OWL. Otherwise you would declare owl:sameAs instead of skos:exactMatch :)

     The more so with broadMatch, closeMatch and narrowMatch which somehow implicitly entail that the concepts matched are indeed different (slightly different, more generic, more specific)

     #1 and #2 can be expressed using convoluted OWL constructions using owl:hasValue restrictions and owl:Nothing ... expression of which is let to the reader as exercise :)

     It would be coool to have an extension of skos enabling direct such declarations, such as

     X skos-plus:noMatchIn V2 (Concept to ConceptScheme)
     X skos-plus:isNoMatchOf Y (Concept to Concept)

     An by the way, those properties could be part of a skos-plus extension including generalization of skos mapping properties to resources which are not skos:Concept, enabling alternatives to the proliferation of abusive owl:sameAs - as discussed with Ivan Herman last week at SemWebPro Paris ... and tweeted by Emmanuelle
     http://twitter.com/#!/figoblog/status/27400161554595840 en français dans le texte :)

     Cheers

     Bernard



     2011/1/27 Emmanuelle Bermes <manue.fig@gmail.com <mailto:manue.fig@gmail.com>>

         Joachim,

         Actually we identified a similar use case at BnF.
         Context is an automated matching or alignment between 2 datasets, that
         has to be repeated on a regular basis (when 1 dataset or the other is
         updated).
         A manual quality check process is set up to check the quality of the
         alignement process.
         A human operator checks that 2 entities that could be automatically
         matched are actually different.
         He wants to record this fact so that in future matchings the manual
         work doesn't have to be done again.

         We plan to use owl:differentFrom to express that those 2 entities are different.
         It seems to me that this case relates to a relationship between two
         URIs in 2 different datasets, rather than a skos:note on one of the
         (un)matched concepts.

         Emmanuelle

         On Fri, Jan 21, 2011 at 4:06 PM, Neubert Joachim <J.Neubert@zbw.eu <mailto:J.Neubert@zbw.eu>> wrote:
          > Hi Bernard,
          >
          > thanks for your answer. I'm happy that I'm not the only one with a
          > requirement for this quite ephemeral feature ;)
          >
          > However, I'm not sure if your solution solves the problem I had in mind: My
          > idea was to express a workflow status. If I got it right, your class NoMatch
          > covers all entities without "Cells" (skos relationships) at a given point in
          > time. But it does not say "for this entity, I have checked intellectually
          > that currently no such relationship can be established".
          >
          > That said, your approach seems quite useful to get hold of the NoMatch
          > entities (and updates automatically if any mapping triple is inserted). I
          > have no experience with reification - is it well supported in your software
          > environment, and does the reification of all skos mapping triples perform
          > well with large vocabularies? It would be very interesting to hear more
          > about the "Terminology Alignment Environment", especially since we have
          > plans to create mappings between different vocabs in the field of economics.
          >
          > Cheers, Joachim
          >
          > ________________________________
          > Von: Bernard Vatant [mailto:bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com>]
          > Gesendet: Freitag, 21. Januar 2011 00:46
          > An: Antoine Isaac
          > Cc: Neubert Joachim; public-xg-lld
          > Betreff: Re: WG: "zero relations" in dataset mappings (has no matching
          > entity in)
          >
          > Minor correction and complement of information.
          >
          > The quoted TAE project correct name is "Thesaurus Alignment Environment".
          > It's currently under development under the OPOCE umbrella, with the
          > technical collaboration of INRIA and Mondeca.
          > There is no public visibility of this project at this point of time, no
          > pointer, sorry ...
          >
          > 2011/1/21 Bernard Vatant <bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com>>
          >>
          >> Hello all
          >>
          >> In the Terminology Alignment Experiment, some applications indeed want to
          >> have this absence of mapping made explicit.
          >> We did it using a subclassing of alignment "Cell", which reifies a skos
          >> mapping (allowing to put metadata on it) between entity1 in source
          >> vocabulary and entity2 in the target vocabulary, in the following way.
          >>
          >> <owl:Class rdf:about="#NoMatch">
          >> <rdfs:label xml:lang="en">No Match</rdfs:label>
          >> <rdfs:subClassOf
          >> rdf:resource="http://knowledgeweb.semanticweb.org/heterogeneity/alignment#Cell"/>
          >> <rdfs:subClassOf>
          >> <owl:Restriction>
          >> <owl:cardinality
          >> rdf:datatype="&xsd;nonNegativeInteger">0</owl:cardinality>
          >> <owl:onProperty
          >> rdf:resource="http://knowledgeweb.semanticweb.org/heterogeneity/alignment#entity2"/>
          >> </owl:Restriction>
          >> </rdfs:subClassOf>
          >> </owl:Class>
          >>
          >> The entity1 in a "NoMatch" cell has no entity2 match whatsoever.
          >>
          >> Maybe convoluted, but saying exactly waht it means.
          >>
          >> Bernard
          >>
          >>
          >>
          >> 2011/1/20 Antoine Isaac <aisaac@few.vu.nl <mailto:aisaac@few.vu.nl>>
          >>>
          >>> Hi Joachim,
          >>>
          >>> No, I've never seen this. It looks in fact a bit odd, as the aligned
          >>> vocabularies may be extended one day so that a mapping can be found.
          >>>
          >>> Re. the representation, there must be ways to express this, using OWL
          >>> class construction mechanisms (your instance of SWD would be in instance of
          >>> the complement class to the class of reosurces that have a SKOS mapping
          >>> property statement with a concept from STW). But I'd be tempted to wait for
          >>> feedback to your questions on the other lists before trying it ;-)
          >>>
          >>> Cheers,
          >>>
          >>> Antoine
          >>>
          >>>
          >>>> Hi,
          >>>>
          >>>> Maybe one of you - from the VocAlign Cluster, especially ;) - has dealt
          >>>> with this?
          >>>>
          >>>> Any hints are appreciated -
          >>>>
          >>>> Cheers, Joachim
          >>>>
          >>>> -----Ursprüngliche Nachricht-----
          >>>> Von: semantic-web-request@w3.org <mailto:semantic-web-request@w3.org> [mailto:semantic-web-request@w3.org <mailto:semantic-web-request@w3.org>] Im
          >>>> Auftrag von Neubert Joachim
          >>>> Gesendet: Donnerstag, 20. Januar 2011 14:11
          >>>> An: Semantic-web@w3.org <mailto:Semantic-web@w3.org>
          >>>> Betreff: "zero relations" in dataset mappings (has no matching entity
          >>>> in)
          >>>>
          >>>> When matching and mapping two datasets, it is common that - on both
          >>>> sides - you find entities which don't have a matching entity on the other
          >>>> side.
          >>>>
          >>>> When that non-matching was verified intellectually, it could be valuable
          >>>> to report this fact - especially to keep track of "false positives"
          >>>> (e.g. matching labels, but different concepts in SKOS systems).
          >>>> Basically, this states a relation between an entity - e.g., a
          >>>> skos:Concept - and a set of entities - as defined e.g. by a
          >>>> skos:ConceptScheme or a void:Dataset.
          >>>>
          >>>> Are you aware of any pattern to express this in RDF?
          >>>>
          >>>> I consider coining something like
          >>>>
          >>>> ext:noMatchingEntity rdfs:subPropertyOf skos:note .
          >>>>
          >>>> Since the date of the above mentioned verification should be reported,
          >>>> you could end up along the lines the following example
          >>>>
          >>>> <http://d-nb.info/gnd/4125416-8> ext:noMatchingEntity
          >>>> [ rdf:value<http://zbw.eu/stw> ;
          >>>> dcterms:modified "2010-01-25"^^xsd:date ] .
          >>>>
          >>>> What do you think?
          >>>>
          >>>> Cheers, Joachim
          >>>>
          >>>>
          >>>
          >>>
          >>
          >>
          >>
          >> --
          >> Bernard Vatant
          >> Senior Consultant
          >> Vocabulary & Data Engineering
          >> Tel: +33 (0) 971 488 459
          >> Mail: bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com>
          >> ----------------------------------------------------
          >> Mondeca
          >> 3, cité Nollez 75018 Paris France
          >> Web: http://www.mondeca.com
          >> Blog: http://mondeca.wordpress.com
          >> ----------------------------------------------------
          >
          >
          >
          > --
          > Bernard Vatant
          > Senior Consultant
          > Vocabulary & Data Engineering
          > Tel: +33 (0) 971 488 459
          > Mail: bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com>
          > ----------------------------------------------------
          > Mondeca
          > 3, cité Nollez 75018 Paris France
          > Web: http://www.mondeca.com
          > Blog: http://mondeca.wordpress.com
          > ----------------------------------------------------
          >




     --
     Bernard Vatant
     Senior Consultant
     Vocabulary & Data Engineering
     Tel: +33 (0) 971 488 459
     Mail: bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com>
     ----------------------------------------------------
     Mondeca
     3, cité Nollez 75018 Paris France
     Web: http://www.mondeca.com
     Blog: http://mondeca.wordpress.com
     ----------------------------------------------------

Received on Friday, 28 January 2011 16:04:16 UTC