Hi Wikidatans,
The conclusion of the 'Primary sorting property' RFC is that property P107
-- main type (GND) -- will be deleted [1, 2, 3]. The most popular option
for replacing P107 is to use properties P31 and P279 ('instance of' and
'subclass of'), which are based on rdf:type and rdfs:subclassOf [4, 5].
Unlike P107, which restricts the world into seven high-level classes,
'instance of' and 'subclass of' give users a flexible way to define an
item's type. They also enable hierarchical classification, and are in line
with Semantic Web conventions.
While general agreement exists on that much, there's active discussion in
the new 'Migrating away from GND main type' RFC [6] about precisely how we
want to capture P107 information with 'instance of' and 'subclass of'.
P107 is Wikidata's most popular property by a significant margin -- it's
currently used in 4,237,061 claims -- so it's important that we get this
right.
Many people on this list are knowledgeable about ontology and
classification but not part of the on-wiki commentariat that tends to
populate Wikidata RFCs. To those people: please give your input in this
migration RFC at
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_….
It has some really interesting questions, like:
A) How should we classify items like 'Ronald Reagan', 'Rabindranath
Tagore'
and 'Coco Chanel'? That is, how should we classify items that have a GND
main type of "person"? A main critique of P107 is that it includes in
'person' things like gods, literary characters, spirits, and individual and
collective pseudonyms. It's clear we want greater precision than P107's
"person", but it's unclear whether it'd be best to map those claims to
instance of 'person', 'human', 'human person', or something else.
Discussion on this is in
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
.
B) How should we classify the GND main types themselves, or the classes
they get mapped to? This is essentially a question about upper ontologies
[7] for Wikidata. There are widely-used third-party upper ontologies like
SUMO, UMBEL and BFO [8, 9, 10]. It's worth discussing which of these -- if
any -- is best to adopt for Wikidata's high-level classification. Although
it probably warrants a separate RFC, some initial discussion of that is in
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
.
C) How can we visualize the type hierarchies we're creating with P31 and
P279? See
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
.
The previous 'Primary sorting property' RFC was easier than this new RFC
will be -- deciding something is a bad idea is easier than actually
replacing it with something better.
So, please join the discussion at
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
!
Thanks,
Eric
https://www.wikidata.org/wiki/User:Emw
1)
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Primary_sorting…
2)
http://lists.wikimedia.org/pipermail/wikidata-l/2013-June/002451.html
3)
https://www.wikidata.org/wiki/Property:P107
4)
http://www.w3.org/TR/rdf-schema/#ch_type
5)
http://www.w3.org/TR/rdf-schema/#ch_subclassof
6)
https://www.wikidata.org/wiki/Wikidata:Requests_for_comment/Migrating_away_…
7)
https://en.wikipedia.org/wiki/Upper_ontology
8)
https://en.wikipedia.org/wiki/Suggested_Upper_Merged_Ontology
9)
https://en.wikipedia.org/wiki/UMBEL
10)
https://en.wikipedia.org/wiki/Basic_Formal_Ontology