RE: HM.Frameworks: Physical Description

[s.livingstone@btinternet.com]

I agree that we need some kind of fuzzy logic. However, in certain cases it would be crazy. It would clearly therefore have to be optional.

I would use it in something like:

<personalInfo>
 <hair Likeness="percent" UpperBound="red" LowerBound="Fair" LikenessVal="76" />
</personalInfo>

and not in something else like:

<personalInfo>
 <age likeness="exact">123</age>
</personalInfo>

These aren't EXACTLY how i would do it - how the heck do we detemine the actualy benchmark for physical chracteristics? I guess sometimes it is relative!! 

<personalInfo>
 <hair Likeness="percent" UpperBound="idref to some hair element" LowerBound="idref to some other hair element" LikenessVal="76" />
</personalInfo>

(actually i just wrote this 4 different ways and can't decide, so i will just leave it the way it is :) - a lot of work to be done) 
But i think some "fuzziness" is required in cases.
Another i would mention happened to me today when I was asked to neter information on a survey. Most of the questions asked me to rate between 1 and 10 - the same kind of fuziess i guess.

<surveyInfo>
 <Q5 Likeness="point" UpperBound="10" LowerBound="1" LikenessVal="6" />
</surveyInfo>


Anyway, just a thought.

Steven Livingstone,
Author Pro XML 2e and others.
http://www.deltabis.com

-----Original Message-----
From: Kurt Cagle [mailto:cagle@olywa.net] 
Sent: Tuesday, August 28, 2001 12:33 PM
To: Bullard, Claude L (Len); humanmarkup-comment@lists.oasis-open.org
Subject: Re: HM.Frameworks: Physical Description

I'm just responding to a thread on this - the KindaLike is Sean's terminology <grin/>.
 
Seriously, all I'm trying to do is look at potential correlative mechanisms based upon specific relational operators. One potential arena where I can see something like this in play is in preference mechanisms, which I see as one specific region in which HumanML has definite applications. As I mentioned before, in some cases these measures are subjective, though I suspect there are relatively few areas where a more formal definition of a given measure couldn't be formulated. Consider color equivalences, for instance, where two reds may be fairly distinct from one another but is more alike than green and very much more alike than blue. If I describe someone as having red hair, for instance, and I want to draw a correlative link to another person with red hair, the like attribute would not likely be 1.0 (I have two daughters with red hair, and yet it is obvious looking at them that one is not the same shade as the others).
 
These don't have immediate bearing to the HumanML schema itself, only to the creation of correlative mechanisms between two instances of that schema.  In that sense, the discussions here are application oriented rather than schema oriented. However, having said that, I would think that correlative measures would make a great deal of sense in Public Safety issues, especially as it brings up another point that has been bothering me for a while. We need to be concerned about the nature of the measurements involved. Should units be introduced into the schema itself, and if so, how? Is it more reasonable to describe a map of units of a given type as a prolog block to any measurement indicators. For instance, it may be more reasonable to basically describe in the schema that a person's height measurement is given as being of type Length, and in the prolog all units of type Length are described as being in centimeters or inches. This makes it possible to make all measurements scalars without the ugly necessities of parsing out unit measures from strings. Note that this may have already been addressed. I'm somewhat backed up in my reading at the moment.
 
-- Kurt