Re: SKOS transitive hierarchical relations

Dear Christophe,
I agree with you since I believe that what you say is not in contrast with
our approach. BTW we do have Information Scientists in our research group
and we do take into account human aspects.

We would treat Turbine Blade as part-of Turbine. Now, we could consider
this part-of as transitive or not. Suppose for a moment to consider it as
non transitive. We categorize non transitive relations as associative, but
this does not mean that they are all the same to us. In fact, we further
specialize hierarchical (transitive) and associative (non transitive)
relations into more fine grained ones [1]. This allows us making the
assumption that transitive closure is computed only on hierarchical
relations.

This distinction allows returning documents about Turbine Blade if query
expansion is requested on the term Turbine (otherwise we only use
relations from the transitive closure). We can parameterize query
expansion according to the kind of associative relations the user/system
wants to take into account with expansion.

Notice also that our knowledge base is organized as a set of faceted
classifications [2] (we actively collaborate with the ISI institute of
Bangalore for this). We might allow Turbine Blade part-of Turbine to be
part of a facet in a given domain (in this case we would enforce the
relation to be transitive), while in contrast we might decide to do not
include Turbine Blade is-a Blade (still transitive but not in that facet
and therefore in specific applications we may decide to treat it
differently).

Cheers,
Enzo

[1] http://eprints.biblio.unitn.it/archive/00002214/01/techRep466.pdf
[2] http://eprints.biblio.unitn.it/archive/00002235/01/techRep468.pdf



> The fact that SKOS is an RDFs application put on the boundaries of OWL
> is sometime creating confusions...
>
> Could I emphasize that there is a need to organize terminology for the
> specific operations done by machines in direct relations with humans?
> OWL is about structured data representation, automated reasoning:
> machine to machine communications.
> SKOS may be part of this (I am not fully involved on that aspect) but it
> is ALSO (mainly in my point of view) a tool for Information Scientists
> (that you call Librarians!) to organize relations between information
> resources and HUMANS (more precisely between Human terms and Machine
> Coded concepts: VoID may be a way to organize relations between
> resources and concepts).
>
> I insist on this distinction and encourage you to discover the world of
> complexities in the man/machine relations for discovery, learning, story
> organization, concept identification, authoring, publication... SKOS is
> part of that. SKOS should always be able to represent a thesaurus (some
> problems remaining). A thesaurus may be built respecting or not the
> rules for an ontology (in its OWL related meaning) but at least SKOS
> respects RDFS rules for information encoding.
>
> For "humans", Broader/Narrower relations are used to indicate "For an
> user asking resources for a given concept, every resources linked to a
> transitiveNarrowers may interest him/her if an "expanded" search is
> requested"
>
> The Turbine Blade example is interesting: it can be ontologically a
> specialization of concepts Turbine and Blade (and one will probably wish
> to specify "how" a Turbine Blade is a Blade and "how" a Turbine Blade is
> a Turbine). But for Information Management, Turbine Blade may be only a
> Narrower of Turbine (If you want to learn about Blades, it may be judged
> that literature about Turbine Blades is not DIRECTLY interesting for
> you; but if you ask for "Turbines", then "Turbine Blades" may be judged
> as always relevant).
>
> It is easier to build a thesaurus using ontologically supported relations.
> But in Medicine and Humanities, it becomes much more difficult:
> http://www.nlm.nih.gov/mesh/meshrels.html
>     (in the MeSH hierarchies, NLM even make one intended BT/NT loop
> between "Moral" and "Ethic" !)
>
> With chemistry, one can organize molecules in different hierarchies:
> * their structure
> * their use
> * their impact on human
> * their different classifications
> etc.
>
> Some of us are fighting to have machines reason better (A.I.), some
> others are worried about humans (Natural Intelligence!).
>
> I think that "Application Profiles" remains an important investigation
> field to improve SKOS and adapt it to different situations.
> What do you think?
>
> Christophe
>
> Le 5/01/2012 09:58, Vincenzo Maltese a écrit :
>> Dear Simon, all,
>> I'm surely not an expert of SKOS, nor a librarian, but at the latest UDC
>> seminar I participated I tried to bring the point of view of my research
>> group (working in knowledge representation and reasoning) in which we
>> tried to formalize the meaning of the BT\NT relations.
>>
>> We interpret them as superset/subset and let them correspond to
>> subsumption in description logic were the interpretation of each node in
>> the vocabulary is the set of documents about the node. Now, since both
>> subset and subsumption are transitive, this works only for transitive
>> BT\NT relations. We treat non transitive BTs as associative relations.
>>
>> +1,
>> Enzo
>>
>>
>>> The problem that with the second version of SKOS (as opposed to the
>>> original, SME-focused, draft that was stable for about four years) is
>>> that
>>> the traditional  semantics of controlled vocabularies as standardized
>>> in
>>> (e.g. ISO 2788) were rejected, without a corresponding recognition that
>>> the
>>> semantics were being changed.
>>>
>>> The traditional domain of interpretation for controlled vocabularies is
>>> #$ConceptualWork; the skos:broader relation in the SME developed
>>> vocabulary
>>> corresponded to BT (Broader Term) in controlled vocabularies.
>>>   Undifferentiated BT is a relationship between subject vocabulary
>>> terms,
>>> and not between #$Collection in an underlying ontological theory.
>>>
>>> BT is sometimes referred to in the literature as The Hierarchical
>>> Relationship; Associative relationships (RT, or skos:related) are the
>>> residual category of relationships that are neither that of equivalence
>>> nor
>>> hierarchical.
>>> The defining characteristic of hierarchical relationships is that they
>>> are
>>> always true - under NISO standards, intensionally, under ISO 2788
>>> extensionally for the domain of documents to which the definition
>>> applies
>>> (illustrated in the acceptability and classification of the the turbine
>>> -
>>> blade example).
>>>
>>> IT was decided that an essential use case for SKOS required that A
>>> broader
>>> B, B broader C  and not A broader C ; it was also decided that the
>>> semantics of broader were not changed by this (a deliberate decision
>>> was
>>> made not to change the namespace). I must confess that I am still not
>>> clear
>>> on how this is possible, or what broader actually means.
>>>
>>> If you want to capture the semantics of Broader Term using the SKOS
>>> vocabulary, you should assert broaderTransitive and ignore the
>>> suggestions
>>> in the primer. UMBEL has been doing the right thing.
>>>
>>> Simon
>>> Simon
>>>
>>
>>
>>
>

Received on Thursday, 5 January 2012 12:40:19 UTC