Re: [Semediawiki-user] Categories and deprecated Relation:Instance_of

[S P. <ski...@ea...>]

Jeff Thompson wrote:
> On this page:
> http://ontoworld.org/wiki/Relation:Instance_of
> it says that Relation:Instance_of and Relation:Subclass_of are deprecated and
> you are only allowed to use the category system.

Whoa, do what you like.  It's only a strong recommendation on ontoworld.org.
That page explains the rationale: "do not use this annotation to 
workaround the faster, more powerful, and better supported category system"
SMW works with MediaWiki's Category system, ambiguous and confused as it 
is, rather than fighting it.

One great benefit is semantic queries follow the transitive nature of 
subcategories, by default up to 10 levels. 
http://ontoworld.org/wiki/Help:Semantic_search says:
"When searching for pages within a category, the result also involves 
all pages that are contained in subcategories of this category."


> Maybe it was decided that...

I don't think anyone made any decision, I think Markus just noticed 
people creating lots of ambiguous confusing "IsA" and "InstanceOf" 
relations on ontoworld, not realizing that MediaWiki's Category does 
something similar.

> average users cannot tell the
> difference between "instance of" and "subclass of", and so we are stuck with
> only "category".  Is that why the decision was made to try to get by with the
> ambiguous use of "category"?

MediaWiki does handle the distinction.  Articles in a category are 
rdf:type (which imples "instance of a class") and subcategories of a 
category are rdfs:subClassOf.  (If you do RDF export of article and 
category pages, you can see this.)  I'm no RDF expert, but this seems a 
reasonable choice.


> For example, here is the Wikipedia category Hindi-language_films:
> http://en.wikipedia.org/wiki/Category:Hindi-language_films
> Note that this category belongs to 3 other categories in 3 very different senses:

You bring up good points about the ambiguity of categories of categories 
in MediaWiki.  This should definitely be noted somewhere in SMW help.

> 1. Subclass of: "Indian_films".  A specific film that is a Hindi language film
>     is also an Indian film, so the category Hindi-language_films is a
>     subclass of Indian_films.

That works.

> 2. Instance of: "Films_by_language".  A specific film that is a Hindi
>     language film is *not* a film by language.  Rather, it is the
>     category Hindi-language_films that is an instance of the higher-level
>     category of Films_by_language, along with other categories like
>     Tamil-language_films.

Yup, from my reading of http://www.w3.org/TR/rdf-schema/#ch_subclassof, 
[[Category:Hindi-language_films]]'s [[Category:Films_by_language]] is 
not an rdfs:subClassOf relation.  I think this would be better as 
rdf:type; you could create a relation, e.g. [[Relation:Category type]], 
that maps to this.

> 3. Somehow related to: "Bollywood".  The category Hindi-language_films has
>     something to do with Bollywood, but the relationship is not specified.

Yes, it seems that this category could instead have a nebulous 
[[Concept::]] relation to an article on Bollywood.  You should then be 
able to use subqueries to find "All articles in any category with a 
Concept relation to Bollywood".  Wikipedia often uses categories as 
browse aids, *because* it doesn't have the fine additional features of 
SMW like Special:Browse and the factbox.

By the way, someone created http://ontoworld.org/wiki/Relation:Related , 
it's also deprecated.


>     However, isn't that the whole point of the Semantic Web, to make the relationships
>     explicit?  But the category system is as generic as an untagged link in
>     Web 1.0 that we are trying to fix with Web 3.0.

If you're using SMW you have the opportunity to tighten up the use of 
Category on your wiki.

> Is there a web page or statement about how the Semantic Mediawiki will
> handle these 3 different types of relations,

I don't think so.  Where would you expect to see it?

> if the more meaningful
> relations are deprecated

Note that "is a" and "instance of" aren't more meaningful.  They're just 
words until you use the special property "Imported from" to relate them 
to a more rigorous ontology.

> and we are only allowed to use the category system?
> Is it a totally closed issue?

It's WORSE than that: an angry mob will descend on your wiki and BURN IT 
TO THE GROUND if spies discover you using SMW in any way counter to 
attempts to guide the uncoordinated random pages on ontoworld.org.
:-)  ;-)

Cheers, thanks for the good discussion.
--
=S Page