Red dogs and blue cats?

[Mason Lee]

I have a question: If I pay attention to only red dogs and blue cats,
how is this going to be represented? Should I have a singular and
fully descriptive uri for each compound-concept, or will APML allow
using RDF or some sort of concept element nesting to refine the
concepts of dogs and cats? To put it another way, is there any
planned support for "adjectives"?

(The rest of this post assumes APML is XML, but it doesn't have to be.)

Here's an pseudo xml example, where you just use a long URI for each:

<concept key="urn:apml-org:blue%20cats" .../>
<concept key="urn:apml-org:red%20dogs" .../>

However, these would explode combinatorially. Some sort of compound
concepts seem better:

Here's psuedo xml example for nesting concepts:

<concept key="urn:ampl-org:cats" ...>
<concept key="urn:apml-org:blue" value=1/>
</concept>
<concept key="urn:ampl-org:dogs" ...>
<concept key="urn:apml-org:red" value=1/>
</concept>

Re: Type Attribute

I'm not a big fan of the proposed "type" attribute for concepts, which
has been suggested would allow for type="Person/Location/Music", etc.
First, enumerating the "types" puts APML right back into defining
ontologies beyond the attention ontology. Second, a technical
problem: if you use an xml attribute for "type", that means that the
concept can only have type, and I couldn't use "type" to be interested
in dogs that are both red and hot.

--Mason

[gdupont]

Hi Mason,

This is an issue : complex or composed concept. What about having a
piece of RDF to detail this concept instead of trying to add
adjectives ?

<rdf:Class rdf:about="http://mason/red_dog">
<rdfs:subClassOf rdf:resource="http://en.wikipedia.org/wiki/Dog"/
>
<ajectiv:color rdf:resource="http://adjectives/color/red"/>
</rdf:Class>

Supposing that you define somewhere the adjectives and colors
ontologies. And thus your APML concept will look like :

<concept key="red dog" uri="http://mason/red_dog">

About the type attribute proposal, I was not thinking to redefine an
ontology, but to provide optional URI to an external ontology of
types.

gd

[scottw]

It might be worth a look at SKOS for this:

http://www.w3.org/TR/2008/WD-skos-primer-20080221/

S

[TSchultz55]

RE:

> It might be worth a look at SKOS for this:
>
> http://www.w3.org/TR/2008/WD-skos-primer-20080221/

SKOS is supported with DBPedia resources.