Re: Finalizing Schema.org "Roles" design

Hi all:

First: I like the proposal and suggest to put it forward.

Second: For conceptual modeling geeks and the group of OncoClean fans: As far as I understand, the "roles" proposal in schema.org is mainly for 

- keeping meta-data about type membership for anti-rigid properties (~R) in OntoClean terminology [1], 
- in such cases where the owner of the data wants to indicate that the type membership is temporarily bound and/or has meta-data about the latter aspect.

Note that in OntoClean, "properties" are not attributes or OWL/RDF/Microdata properties, but unary predicates, e.g. in the sense of type membership. In the roles proposal, this is further applied to both unary predicates (in the form of type membership) and binary predicates (in the sense of "object properties").

Being a football player is something one can become and stop being without ceasing one's existence. But a Person can, to the degree of our knowledge, not turn into a boombox, and a car will unlikely morph into an integer number. Note that this does not imply that individuals of a given type could not cease to exist. It is just about the possibility that type membership can be acquired or lost independent of the existence of the individual.

Your place of birth is a rigid property, your place of residence is anti-rigid.

For rigid unary properties, one can add the meta-data directly to the entity representing the individual (e.g. the date of becoming a person is the date of birth of this person). 

For rigid binary properties, there is typically little need for additional meta-data related to that property which cannot be as well represented using another property for the main entity. 

Martin

[1] http://en.wikipedia.org/wiki/OntoClean#MetaProperties


On 09 May 2014, at 02:53, Thad Guidry <thadguidry@gmail.com> wrote:

> @Niklas,
> 
> I think it should be that a "conditioned", "variable", or "temporal" use that determines the primary use of "Role".  "variable" is a primary motivator.  In Freebase we have the idea of CVTs or Mediators, which are sub-graphs, if you will, inside a single property...which is kinda what this whole Roles proposal is hinting towards, and for good reason.  There's a real need.
> 
> I do wonder if the term "Role" should perhaps be changed to that of "Temporal" or "Variable" or as we chose in Freebase the term, "Mediated".
> 
> Dan mentioned in my comment on the doc proposal that this is an evolving thing.  But we do need a super-type like Role to evolve around as Martin is asking.
> 
> @Jarno,
> Keep those awesome playground constructs coming...the more we see the potential use / mis-use of Roles, the better for this discussion going forward.
> 
> @Dan,
> I am happy with the proposal thus far, as it is.  It fills the gap, will lead to confusion for some, but there is no medicine to cure the confusion other than perhaps asking folks to use/abuse it and work in real examples and experiments, as Jarno loves doing.  But what about using the term "Mediated" instead of "Role", to give it longer life, and broader usage as a super-type ? Slippery slope approaching, I know. But not that slippery.
> . 
> -- 
> -Thad
> +ThadGuidry
> Thad on LinkedIn

Received on Friday, 9 May 2014 08:45:37 UTC