Re: SemWeb terminology page

Hi everyone,

Before it's too late, I'd like to remind that we discussed that yesterday in cologne [1] and came with the following proposals (not a definitive choice!) which Jodi may like :-)
by the way Mark's mail was really helpful to guide our discussion!

Antoine

http://lists.w3.org/Archives/Public/public-xg-lld/2010Dec/0005.html

=====
Based on Mark's mail: http://lists.w3.org/Archives/Public/public-lld/2010Nov/0121.html

Terminology choices should be guided by (in that order)
1. Ease of understanding for librarians
2. Compatibility with LD technical terminology
Also, we can't do without a definition and example of resources that are included in the groups.

We found some terms easier to rule out for various reasons: model, list, codes, SKOS, schema, terms.

For group 1, among the proposed expressions "value vocabulary" is a (relative) best.
Group 1 should include also person lists (VIAF), places, events. Certainly also reference sets of works (like Getty's CONA). In fact group 1 can contain pretty much anything, except what falls in Group 2 :-) It's hard to think of a dataset that couldn't possibly be used in group 1 one day. Belonging to that group appears to be based on usage, not on essential characteristics of what's described. In fact we are very tempted by using the word "dataset". We could do not really make a choice between:
- "dataset" alone--on the condition that we explain the specific role that can be played by certain library datasets such as thesauri, authority list, etc.
- "reference dataset"
- "organized dataset"

For group 2, everyone agrees that this is the set of resources that in the RDF world will be expressed as "RDF vocabularies" (as defined in the RDF schema spec), namely, classes and properties. However, using "classes and properties" straight away may be hard as a first term for librarians. Further, choosing that may result in ruling out these vocabularies/models/schemes that are not (yet) implemented as RDF vocabularies.
A proposal that did not meet strong objections is to us "metadata element set" and say in the definition that this will correspond in RDF to classes and properties.
=====





> The use of 'vocabulary' with different modifiers seems doomed to fail. That's because, for me, I find it difficult to mentally distinguish 'Value vocabulary' and 'element vocabulary'. The idea of a 'library vocabulary' and 'semantic web vocabulary' is just barely understandable enough for me to handle.
>
> I'd be very, very happy if someone could propose an alternative which didn't use 'vocabulary' twice. I fear abbreviation as well as the assumption that, oh, yeah, we know what vocabularies are (with different resultant assumptions depending on one's background).
>
> I think it's important to note that FRBR is used to refer both to ontologies and to a conceptual model. It's worth explaining the difference between an ontology (completely formally specified) and a conceptual model, if we can.
>
> -Jodi
>
> On 2 Dec 2010, at 21:43, Tillett, Barbara wrote:
>
>> I agreee with Tom:
>>     Value vocabulary
>>     Element vocabulary
>>     Data set
>>     No particular handle for "conceptual models", which we
>>         may refer to variously as "ontologies" or "conceptual
>>         models".
>> It fits with what has been common parlance when talking about these in library communities.- Barbara Tillett
>>
>> -----Original Message-----
>> From: public-lld-request@w3.org [mailto:public-lld-request@w3.org] On Behalf Of Thomas Baker
>> Sent: Thursday, December 02, 2010 4:30 PM
>> To: Mark van Assem
>> Cc: public-lld
>> Subject: Re: SemWeb terminology page
>>
>> On Tue, Nov 30, 2010 at 02:47:14PM +0100, Mark van Assem wrote:
>>> We from the DO cluster had another discussion about a terminology issue.
>>>
>>> ----------------------------------------------------------------------
>>> To reiterate, the issue is what term to apply to two groups of things
>>>
>>> 1) LCSH, AAT, WordNet and the like. These describe concepts that are
>>> used in actual medata.
>>>
>>> 2) FOAF, FRBR and the like. These describe what concrete metadata must
>>> look like; defines classes and properties, the instances of which are
>>> actual metadata, and in which concepts defined in 1 are used.
>>> ----------------------------------------------------------------------
>>>
>>> To resolve this we thought it's useful to define criteria:
>>>
>>> a) the terminology should fit with our (main) public. This is probably
>>> library people, so how well it fits with the SemWeb folks is secondary.
>>>
>>> b) because especially the word "vocabulary" is confusing, we should
>>> either avoid "vocabulary" altogether or prepend it so as to
>>> distinguish the two e.g. "value vocabulary" and "metadata vocabulary"
>>>
>>> c) given that people tend to abbreviate terms when they use them, the
>>> prepending approach may create the confusion we're trying to avoid.
>>>
>>> d) if possible it is advantageous to not refer to how the groups are
>>> implemented, i.e. do not refer to RDF, XML or anything else.
>>> Definition through mentioning how it is implemented can be distracting.
>>>
>>> ----------------------------------------------------------------------
>>> I've collected all proposals and added a few more:
>>>
>>> Suggested terms for group 1:
>>> - vocabulary
>>> - value vocabulary
>>> - SKOS vocabulary
>>> - KOS
>>> - domain vocabulary
>>> - controlled list
>>> - code list
>>> - "thesauri, glossaries, classification schemes and other vocabularies"
>>
>> +1 for "value vocabulary" as long as we include the disclaimer
>> that the vocabulary is being characterized as a "value"
>> vocabulary because that is how its terms are typically used.
>> I agree with Mark and Mikael on this.
>>
>>> Suggested terms for group 2:
>>> - RDF vocabulary
>>> - properties / property set
>>> - Set of property and class terms
>>> - metadata vocabulary
>>> - data elements
>>> - element vocabulary
>>> - ontology
>>> - conceptual model
>>> - metadata element set
>>> - metadata model
>>> - metadata schema
>>> - modelling schema
>>
>> I prefer Mikael's imperfect "element vocabulary", which seems approximately right if you squint hard enough.  Again, one would need to qualify this by saying that the vocabulary is being characterized this way because they largely consist of properties, which are typically used as predicates.
>>
>>> ----------------------------------------------------------------------
>>>
>>> Mikael's suggestion of "value vocabulary" and "element vocabulary" is
>>> really good, but this does mean we'll have to be very careful in
>>> drafting our documents to meet criterium c, but this will not help
>>> people outside our documents. We have a chance here to think up
>>> something that will have a wider use than our documents.
>>>
>>> If we ignore criterium d, then choosing "RDF vocabulary" for 2 seems
>>> obvious, but violates criterium c. Therefore something like
>>> "conceptual model" or "ontology" for group 2 is better.
>>>
>>> My personal favorites are:
>>>
>>> 1) value vocabulary - expresses that we're dealing with vocabulary
>>> concepts that are _used_ in actual metadata. Violates b/c though.
>>>
>>> 2) metadata model - expresses that these things determine how actual
>>> metadata can look like; more neutral term than "ontology" which
>>> library people may interpret as something different than a metamodel.
>>
>> I find "metadata model" really confusing!  For example, is an XML schema a "metadata model" (because it expresses how actual metadata can look)?
>>
>> Given a choice, I would rather bite the bullet and use "ontology", taking the risk that some readers may still associate "ontology" indelibly with a branch of traditional metaphysics. Or "conceptual model".  But I do not see these as being in the same category as "metadata element set".
>>
>> On the other hand, I'm not really sure we need to define a separate handle for things like FRBR.  To take the example of SKOS, SKOS is definitely an RDF vocabulary -- in our terminology, an "element vocabulary" (with classes that might be considered a "value vocabulary", but basically an element vocabulary) -- but it is also a data model for describing concepts.  For our purposes, if we squint hard enough, I think we could characterize SKOS as an "element vocabulary" and simply mention that it is related to a data model -- without classifying it as a "metadata model".
>> The same applies to FRBR: yes, it is a conceptual model, but when used in linked data it is an "element vocabulary".
>>
>> This is the best I can come up with given the choices, but none of the options seems really satisfactory.  I only wish I were more firmly convinced that this exercise of "haute vulgarisation" is really helpful and will not itself cause some confusion...
>>
>> +1 for Ed's suggestion re: "dataset".
>>
>> In sum:
>>     Element vocabulary
>>     Value vocabulary
>>     Data set
>>     No particular handle for "conceptual models", which we
>>         may refer to variously as "ontologies" or "conceptual
>>         models".
>>
>> Tom
>>
>>> Opinions? If you argue pro/con particular terms it helps if you point
>>> out the principles/communities on which that preference is based.
>>>
>>> Best,
>>> Mark.
>>
>> --
>> Tom Baker<tbaker@tbaker.de>
>>
>>
>
>

Received on Thursday, 2 December 2010 22:23:53 UTC