AW: Predicate for quasi-synoyms (with upwards posting)?

Dear colleagues,

Thank you all for your feedback! I've answerd to some mails, but unfortunately not included the list, so I'll try to summarize the suggestions, before introducing my solution:


"upwards posting" not considered a good practice / why not define seperate concepts

As mentioned in many answers, the whole "upwards posting" looks a bit messy. I found it already in place when I started working with STW some years ago. However: For every subject vocabulary, there are peripheral areas, where it makes sense to work on a coarser granularity of concepts. For the example given below ("Social class") the STW defines explicit narrower concepts (see http://zbw.eu/stw/descriptor/15717-1), whereas a taxonomy for mathematics could choose to do not so. This is in accordance with ISO 25964, 8.4. A valid reason mentioned there is to reduce the number of concepts in a vocabulary, in particular if very little information on the more specific concepts would be available in the given context. 

Nevertheless, these terms exist. Often, explicit descisions have been made about their subsumption under a broader concept, and in my eyes it makes sense to document these descisions and make the broader concept retrievable via the subsummarized term (particularly for indexing, as Stella notes). Even in selected retrieval/search expansion cases these terms may be useful (think of building categorial facets for search results of indexed and non-indexed documents), while in most cases you definitively want to exclude them.


use skos:hiddenLabel

I've considered a usage similar to what Christophe suggested, but ended up with dedicating skos:hiddenLabel to _invisible_ labels (mis-spellings and the like). Because of the documentary value discussed above, I want to show the "upwards-posted" terms to the thesaurus users, e.g. like this

Social class
                used for: Social background
                also utilized for: Elite, Middle class, Working class


use SKOS-XL labels, or lexinfo:approximateSynonym/lexinfo:hyponym

Since these are object properties, they would require separate URIs. That would mean "more things" in the concept scheme, which I wanted to avoid in the first place - plus increased structural complexity :-(


Since it looks like nobody has already created what we search for, I've introduced a custom specialization of skos:altLabel

<http://zbw.eu/namespaces/zbw-extensions/altLabelNarrower>
    a owl:AnnotationProperty ;
    rdfs:isDefinedBy <http://zbw.eu/namespaces/zbw-extensions/zbw-extensions.rdf> ;
    rdfs:label "altLabelNarrower"@en ;
    rdfs:subPropertyOf skos:altLabel ;
    skos:definition """Lexical label for a narrower term which is
    not defined as a separate concept (according to ISO 25964, 8.4), sometimes
    referred to as 'upward posting'."""@en .

This is not carved in stone, so I'd be happy to receive further feedback.

Thanks again, and have fine holidays - Joachim

PS. I'm aware that the specialization above does not cover the use case illustrated by Christophe with the "Coke/Cola/Cocaine" example. However, if additional functionality is required (asking the use to choose between alternatives), a separate property would make sense to trigger such a functionality explicitely.


> -----Ursprüngliche Nachricht-----
> Von: Stella Dextre Clarke [mailto:stella@lukehouse.org]
> Gesendet: Freitag, 31. März 2017 21:33
> An: Antoine Isaac; public-esw-thes@w3.org
> Betreff: Re: Predicate for quasi-synoyms (with upwards posting)?
> 
> Upward posting may have some useful contexts at the indexing stage, but not
> for automatic search expansion, I suggest. It can be a useful source of ideas in
> contexts where a human intermediary chooses to increase recall, but not
> where the machine applies it blindly.
> 
> Stella
> 
> 
> On 31/03/2017 19:20, Antoine Isaac wrote:
> > Hi Joachim, all,
> >
> > I don't know of any predicate that would answer the need. Maybe
> > because upward posting is not really recognized as a good practice in
> > general ;-)
> >
> > Antoine
> >
> > On 28/03/17 10:13, Thomas Francart wrote:
> >> Not exactly a specialization of skos:altLabel, but Lexinfo defines
> >> http://www.lexinfo.net/ontology/2.0/lexinfo#approximateSynonym or
> >> http://www.lexinfo.net/ontology/2.0/lexinfo#hyponym to be used to
> >> relate two LexicalSenses in Ontolex.
> >> Why aren't "Elite", "Middle class" and "Working class" defined as
> >> skos:narrowers of "Social class" ?
> >>
> >> Thomas
> >>
> >> 2017-03-28 9:51 GMT+02:00 Neubert, Joachim <J.Neubert@zbw.eu
> >> <mailto:J.Neubert@zbw.eu>>:
> >>
> >>     SKOS does not distinguish between synonyms and quasi-synonyms
> >> resultig from „upwards posting“ (as mentioned in
> >> https://www.w3.org/TR/skos-primer/#secalt

> >> <https://www.w3.org/TR/skos-primer/#secalt>). To give an example
> >> (taken from
> >> https://www.slideshare.net/accessinnovations/taxonomy-fundamentals-wo

> >> rkshop-2013/37
> >> <https://www.slideshare.net/accessinnovations/taxonomy-fundamentals-w

> >> orkshop-2013/37>):____
> >>
> >>
> >>     __ __
> >>
> >>     :xyz skos:prefLabel “Social class” ;____
> >>
> >>          skos:altLabel “Elite”, “Middle class”, “Working class” .____
> >>
> >>     __ __
> >>
> >>     When in a retrievel or voculary mapping application each pref-
> >> and altLabel is used as synoym for each other, this creates nasty
> >> problems: If, e.g., the user searches for “Elite”, she also gets
> >> documents with “Middle class” and “Working class”.____
> >>
> >>     __ __
> >>
> >>     We consider introducing some custom sub-property of skos:altLabel
> >> to differentiate the cases, like this:____
> >>
> >>     __ __
> >>
> >>     :xyz skos:prefLabel “Social class” ;____
> >>
> >>          skos:altLabel “Social background” ;____
> >>
> >>          custom:altLabelNarrower “Elite”, “Middle class”, “Working
> >> class” .____
> >>
> >>     __ __
> >>
> >>     Is anybody aware of custom specializations of skos:altLabel for
> >> quasi-synonyms we could re-use?____
> >>
> >>     __ __
> >>
> >>     Any hints appreciated – cheers, Joachim____
> >>
> >>     __ __
> >>
> >>
> >>
> >>
> >> --
> >> *
> >> *
> >> *Thomas Francart* -*SPARNA*
> >> Web de _données_ | Architecture de l'_information_ | Accès aux
> >> _connaissances_ blog : blog.sparna.fr <http://blog.sparna.fr>, site :
> >> sparna.fr <http://sparna.fr>, linkedin :
> >> fr.linkedin.com/in/thomasfrancart
> >> <https://fr.linkedin.com/in/thomasfrancart>
> >> tel :  +33 (0)6.71.11.25.97, skype : francartthomas
> >
> 
> --
> *****************************************************
> Stella Dextre Clarke
> Vice-Chair, ISKO UK and Vice-President, ISKO Luke House, West Hendred,
> Wantage, OX12 8RR, UK
> Tel: 01235-833-298
> stella@lukehouse.org
> *****************************************************
>

Received on Wednesday, 12 April 2017 15:47:33 UTC