[Data-modeling] Physical object

[Danny Hillis]

I am sympathetic to what you are striving for, but I think Kurt is  
speaking from hard-won experience.  When we first started data  
modeling in Freebase we tried to make types too rigorous,  
hierarchical, orthogonal. What we discovered is that when we made them  
that way, they were often unintuitive to the people who had to use them.
-Danny


On May 4, 2008, at 7:03 PM, Ed Laurent wrote:

> Thanks for the great response. You touched on all the subjects and  
> examples I was thinking about. I'm just thinking of ways to build  
> intuitive models that also have some of these rigorous,  
> hierarchical, orthogonal qualities. Sounds like a physical object  
> type is not the way to go.
>
> -Ed
>
>
> On Sun, May 4, 2008 at 9:41 PM, Kurt Bollacker <kurt at metaweb.com>  
> wrote:
>
> On Sun, May 04, 2008 at 01:39:10PM -0400, Ed Laurent wrote:
> > Does anyone know of an existing data model for a generic physical
> > body<http://www.freebase.com/view/en/physical_body>/object
> > that has volume, mass, and maybe some shape properties (e.g., max  
> length)? I
> > can't find any that aren't specific to a particular class of object.
>
> I don't believe there is a physical object type, but I'm interested in
> what you'd be doing with it.
>
> Most (All?) topics types in Freebase have been modeled so that the
> properties reflect the usage of that object.  The usage of "physical
> object" is very abstract and ptobably requires a lot of cotypes to be
> of much use.  Systems that have such rigorous, (usually hierarchical)
> orthogonal ontologies (e.g. Cyc) are often designed for automated
> reasoning. Freebase, on the other hand, uses "mix-ins" of cotypes to
> organize information in a immediately, intuitively useful way, with
> evolves toward rigor over time as a goal.  For example, in Freebase,
> the radius of a star is given instead of volume, because that is a
> more accepted measure.  The volume of a person is not given at all
> because it is rarely used by anyone.  The volume of a "body of water"
> is given as cubic kilometers, and the volume of a digital camera is
> represented as its own "cameria dimensions" type.  Each of these is
> tuned to how it is expected to be used.
>
> Perhaps a "phsyical object" type is a good idea, but it's worth having
> the discussion of how to fit it in the ecosystems of existing types
> and needs.
>
>                                                                 
> Kurt :-)
>
>
>
> > -Ed
>
> > _______________________________________________
> > Data-modeling mailing list
> > Data-modeling at freebase.com
> > http://lists.freebase.com/mailman/listinfo/data-modeling
>
> _______________________________________________
> Data-modeling mailing list
> Data-modeling at freebase.com
> http://lists.freebase.com/mailman/listinfo/data-modeling
>
> _______________________________________________
> Data-modeling mailing list
> Data-modeling at freebase.com
> http://lists.freebase.com/mailman/listinfo/data-modeling