[SKOS] "Mapping" vs "standard" relationships

All,

To keep momentum from the call...

The Primer [1], section 3.1, currently says:

      By convention, mapping properties are used to
      represent links that have the same intended meaning
      as the "standard" semantic properties, but with a
      different application scope. One might say that mapping
      relationships are less <em>inherent</em> to the meaning
      of the concepts they involve. From the point of view
      of the original designer of a mapped KOS, they might
      even sometimes be wrong.

      Mapping properties are expected to be useful in specific
      applications that use multiple, conceptually overlapping
      KOSs. By convention, mapping relationships are expected
      to be asserted between concepts that belong to different
      concept schemes. However, the use of mapping properties
      might also be appropriate in cases where someone other
      than its owner needs to enrich the semantic relationships
      within a particular concept scheme.

Alistair suggested dropping the first paragraph above [2]:

    >In fact, I would be tempted drop the first of these three paragraphs
    >altogether. If I had no prior knowledge of SKOS, I would find the
    >first two sentences ambiguous. The words "scope" and "inherent" are
    >particularly difficult here.  And I'm not sure what value the third
    >sentence adds. I.e. one hopes that cases where the KOS designer and
    >the KOS mapper completely disagree about the nature of a mapping link
    >would be very rare. A brief, casual mention such as this may leave the
    >wrong impression, e.g. that these cases could be quite frequent.

to which Antoine responded [3]:

    > In fact I expect that these cases would be quite frequent. If a KOS 
    > designer agreed that a mapping link between two concepts in her KOS fit her 
    > intent when creating the KOS, she would have created it as a standard 
    > semantic relationship then, wouldn't she?

In Washington, we resolved [4]:

    RESOLUTION: 1. keep the mapping vocabulary broadMatch,
    narrowMatch, 2. broadMatch, narrowMatch, etc. are
    rdfs:subPropertyOf broader, narrower, 3. there are
    no semantic conditions on broadMatch, narrowMatch;
    i.e. graphs 1-6 are all consistent, 4. there is some text
    about cultural conventions explaining where we expect
    broadMatch to be used, 5. by convention, mapping properties
    are only used to link concepts in different schemes, 6. in
    the Last Call WD we'll note that the mapping vocabulary
    may be dropped

The discussion in Washington referred to four scenarios: KOS
Description, KOS Mapping, KOS Extension, and KOS Enrichment
[5].

SKOS Reference [6] currently says:

    SKOS Reference: The mapping properties skos:broadMatch,
    skos:narrowMatch and skos:relatedMatch are provided as
    a convenience, for situations where the provenance of
    data is known, and it is useful to be able to tell "at
    a glance" the difference between internal links within
    a concept scheme and mapping links between concept schemes.

The position in SKOS Reference is in line with the Washington
resolution, and I understood Alistair to say we should make
no further commitment beyond this and let the difference
between the mapping and standard relationships sort itself out
in practice.  

I do not disagree, and I think SKOS Reference is appropriately
vague on this issue.  As per point 4 of the resolution above,
however, I think the Primer should provide a bit more guidance
on where we expect the mapping properties to be used.

I understand the Primer example above [1] to mean that given
concept scheme A with concepts X and Y, a person other than the
designer of concept scheme A who wants to assert relationships
between X and Y would use the _mapping_ properties [3].  (I see
that in [5], Alistair thought it might be more appropriate to
use semantic relation properties for KOS enrichment.)  If so,
then the essential difference between mapping and standard
relationships would not seem to lie with whether or not the
concepts are in the same scheme, but rather with the position
of the person making the assertion with respect to the scheme.
Hence my suggestion [7]:

> The argument could run as follows: Ideally, we should be able
> to tell from provenance information who said what, but in
> practice, Semantic Web data is often merged in simple ways that
> obscure the origins of assertions.  The distinction between
> "mapping" and "standard" relationships is one of etiquette --
> directly asserting "standard" relationships sends the message
> that the asserter considers herself qualified to define the
> relationship in a standard way.  For everyone else, the polite
> thing is to assert a "mapping" relationship.

If the criterion for using mapping as opposed to standard
links has to do with whether the related properties are in the
"same" or "different" schemes, I'm not sure what that means
in situations of KOS enrichment (see above) or KOS evolution
(where I may map to another concept using a mapping property,
then later decide to incorporate it into "my" scheme).

I do not think that defining the difference in terms of
standpoint of the asserter is incompatible with the minimal
commitment made in SKOS Reference [6], but it seems easier
to explain.  I think the Primer text should, at any rate,
echo the point in SKOS Reference about "situations where the
provenance of data is unknown".

Tom

[1] http://www.w3.org/2006/07/SWD/SKOS/primer/primer-20090113.html
[2] http://lists.w3.org/Archives/Public/public-swd-wg/2009Jan/0033.html
[3] http://lists.w3.org/Archives/Public/public-swd-wg/2009Jan/0037.html
[4] http://www.w3.org/2008/05/06-swd-minutes.html
[5] http://www.w3.org/2006/07/SWD/wiki/WashingtonAgenda/MappingIssues
[6] http://www.w3.org/2006/07/SWD/SKOS/reference/20081001/
[7] http://lists.w3.org/Archives/Public/public-swd-wg/2009Jan/0039.html

-- 
Tom Baker <tbaker@tbaker.de>

Received on Tuesday, 13 January 2009 22:24:43 UTC