Dear all,
Picking up on an interesting thread from Paul Houle, Renato
Iannella, and Karen Coyle in December [1,2,3,4,5]:
-- Consumers and producers of linked data are faced with
"multiple ways to say things" -- a proliferation of
properties with seemingly overlapping scope, such as
dcterms:subject, sioc:topic, the (draft-only) skos:subject,
and foaf:topic and foaf:primaryTopic. In the absence of
an official relationship between such terms, linked data
users are left to decide for their own how to interpret
the apparent overlap. For producers to express particular
statements in every reasonable alternative variant would
lead to an exponential explosion of triples.
-- Karen points out that the difficulty lies in
understanding the exact meaning of a metadata term given
what are often terse or vague definitions (e.g., the example
of dcterms:title, dcterms:title as annotated by BIBO, and
foaf:name) and wonders where the solution lies -- in better
definitions? more contextual description? One practical
solution lies in creating locally constrained subproperties
of generalized properties, but where on the Web can one
efficiently find an overview of reusable metadata terms?
-- Knowing the domain of a property (e.g., foaf:Document --
and not foaf:Person -- for foaf:primaryTopic) merely shifts
the problem to that of understanding how classes of things
in the world relate to each other. While a set of relevant
things in the world may reflect consensus in a particular
community, the diversity of worldviews in the "open
world" inevitably confound attempts to "close the world".
One may conclude from this:
-- Understanding the meaning and applicability of properties
and classes, and how they relate to properties and classes
in other vocabularies, requires intellectual interpretive
effort.
-- It would spare individual users much ad-hoc duplication of
effort if alignment claims were captured formally and
available for use by all. While alignments can in principle
be declared by anyone, it would be particularly helpful to
know how equivalences between vocabularies are interpreted
by their owners.
Ideally, maintainers should agree on alignments between
their own vocabularies and make reciprocal claims. The FOAF
Project and DCMI made a first step in this direction last
October by mutually declaring the equivalence of foaf:maker
and dcterms:creator -- a claim that is now recorded in the
specifications and machine-readable schemas of both FOAF
and DCMI Metadata Terms.
Shifting vocabulary maintenance effort towards alignment
would address the shortcomings of terse or vague definitions
by putting terms into a richer context, as users uncertain
how to apply dcterms:creator might look to foaf:maker for
additional usage hints, and vice versa.
Alignments are a first step in the direction outlined by
Mike Bergman in his keynote at DC-2010 in Pittsburgh [6].
They are part of the challenge to be addressed at the DC-2011
conference, 21-23 September in The Hague, under the banner of
"Metadata Harmonization: Bridging Languages of Description"
[7].
Tom
[1] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-GENERAL;6377e6f.1012
[2] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-GENERAL;9931d058.1012
[3] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-GENERAL;5200dbe2.1012
[4] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-GENERAL;74ab69da.1012
[5] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=DC-GENERAL;e3e3d45f.1012
[6] http://dublincore.org/workshops/dc2010/DC-2010_20101022_Bergman_keynote.pdf
[7] http://dcevents.dublincore.org/index.php/IntConf/dc-2011/
--
Thomas Baker <[log in to unmask]>
|