For "a green coat" I would attach a "green" attribute descriptor directly
to the "coat" object descriptor as an attribute. For, "The coat is green,"
I would create an "is" event descriptor and make the "coat" object
descriptor its subject and the "green" attribute descriptor its object.
This is functionally the same as attaching the attribute descriptor
directly to the object descriptor, except that additional timing
information is conveyed. Keep in mind that you can also say, "The coat *was*
green," or, "The coat *will be* green." Most people seem to overlook this
when designing their conceptual models.
Don't ignore the temporal and other meta aspects of "static" knowledge. Set
membership can't represent this for you, nor can other simple mathematical
relations. The problem is that mathematical relationships can't be treated
as first-class objects, and cannot be modified with additional
information. In the standard mathematical treatment, you can assert or deny
a statement, but you don't refer to the statement as an entity in its own
right. You can refer to set membership as a mathematical construct in its
own right, but you can't refer to the membership of a particular object to
a particular set as a mathematical construct, nor can you provide
additional details of that relationship on an ad hoc basis as they become
available. How, in your system, are you going to talk *about *the coat
being green, and relate that to other facts and events? How are you going
to apply time constraints ("It was green after I dyed it"), or say just how
green the coat was, or what shade of green, or what quality of green
(mottled, neon, dull, faded, etc.), or how certain you are/were of any of
this additional information at any point in time?
Set theory is grossly inadequate. You need a descriptive system that is
capable of under- or over-specifying the current state and past/future
history of reality, including talking about descriptions (and elements
thereof) expressed in that system. You should be able to encode whatever
information is available, regardless of its nature, and build an
arbitrarily complex model or simulation of any scenario. You need to be
able to do things like this:
*my_coat = ENTITY()*
*my_coat.kind = KIND("coat")*
*my_coat.owner = speaker*
*mottled = ATTRIBUTE()*
*mottled.kind = KIND("mottled")*
*mottled_green = ATTRIBUTE()*
*mottled_green**.kind = KIND(**"green")*
*mottled_green**.attributes.add(mottled**)*
*my_coat_is_mottled_green = COPULA()*
*my_coat_is_mottled_green.**kind = KIND("be")*
*my_coat_is_mottled_green.**subject = my_coat*
*my_coat_is_mottled_green.attribution** = mottled_green*
*my_coat_is_mottled_green.certainty = .95*
*my_coat_is_mottled_green.times.add(now)*
*I_believe_my_coat_is_mottled_green = EVENT()*
*I_believe_my_coat_is_mottled_green.**kind = KIND(**"believe")*
*I_believe_my_coat_is_mottled_green.**subject = speaker*
*I_believe_my_coat_is_mottled_green.**direct_object =
my_coat_is_mottled_green*
*I_believe_my_coat_is_mottled_green.certainty = .99*
*I_believe_my_coat_is_mottled_green.times.add(now)*
This is an oversimplification, of course, but I hope it gets my point
across nonetheless. Any system that deals with raw, static mathematical
assertions is going to be insufficient. The system must be capable of
change, it must be capable of self-reference, and it must be capable of
arbitrary expressiveness, or it will ultimately be nothing more than a toy
system.
On Sat, Jan 24, 2015 at 10:59 AM, Piaget Modeler via AGI <[email protected]>
wrote:
> For the NLP crowd.
>
> How do you represent *IS* ? Do you differentiate *IS* from *TYPE-OF *(i.e.,
> *IS-A*), or *INSTANCE-OF* ?
>
> Take for example,
>
> IS(apple, fruit) - TYPE-OF
> IS(John_Smith, Politician) - INSTANCE-OF
> IS(my_coat, green) - ???
>
>
> Cyc uses individuals and collection and treats generalization as set
> membership between the instance
> and the collection. They use Genls (generalizes) rather than IS-A. But
> I don't recall how they represent
> IS.
>
> How do you represent these distinctions?
>
> ~PM
>
>
> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now>
> <https://www.listbox.com/member/archive/rss/303/23050605-2da819ff> |
> Modify
> <https://www.listbox.com/member/?&;>
> Your Subscription <http://www.listbox.com>
>
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
