Re: Inheritance and RIF

From: "Hans Teijgeler" <hans.teijgeler@quicknet.nl>
Subject: Inheritance and RIF
Date: Sun, 2 Apr 2006 21:58:50 +0200

> Folks,
>  
> I looked for the word "inheritance" in the W3C website, but didn't find
> what I was looking for.
>  
> Assume we have a class called 'lightweight boxer'.
> The "criteria for membership" for that class are, amongst others, that
> the weight of any male boxer of that class, right before a match, shall
> be between 130 and 135 lbs. So I may introduce the Properties minWeight
> and maxWeight, with these values.

Hmm, well, here you are making one of the standard modelling choices that
can lead to problems.  

Note well that the meaning of the "properties" minWeight and maxWeight are
very different from the property weight.  The former two are, perhaps,
properties of the class whereas the latter one is a property of instances
of the class.

> Now I have John Doe, and I type him as 'lightweight boxer'. 
>  
> Are now the minWeight and maxWeight Properties inherited, even where
> these are meaningless for a person? (because an individual doesn't have,
> at any given point in time, a range value, but a point value)

And here is where your modelling choice bites you.

> Or is inheritance only a function that software (reasoners and the like)
> must handle? 

Well, "inheritance" is, at best, a description of what goes on in some
relationships between classes and their instances, or classes and their
subclasses.  It is far better to think logically, i.e., *what* follows (or
should follow) instead of *how* to make it follow.

> Will RIF become a suitable mechanism for that? In this case we would
> need a function that looks at the weight of John Doe and checks whether
> or not that is between minWeight and maxWeight, and if yes confirms that
> the asserted typing was correct, or would even generate that typing (of
> course including other ciriteria, otherwise we would have zillions of
> lightweight boxers in the world).

Well, RIF shouldn't have any mechanism for inheritance.   After all, RIF is
not about classes and instances, but is instead about rules.

> By the way: does anyone know how to represent a value with a tolerance
> in RDF/OWL? So something like: diameter = 150 mm +0.03%  -0.25%.
>  
> Regards,
> Hans

My suggestion to you would be to read the KR literature, where this entire
issue was discussed at length.  A good starting point would be 

 title=		"What's in a Link:
		 Foundations for Semantic Networks",
 author=	"Woods, William A.",
 pages=		"35--82",
 booktitle=	"Representation and Understanding:
		 Studies in Cognitive Science",
 editor=	"Daniel G. Bobrow and Alan M. Collins",
 year=		1975,
 publisher=	"Academic Press",

This paper also appears in:

 editor=	{Ronald J. Brachman Hector J. Levesque},
 title=		"Readings in Knowledge Representation and Reasoning",
 booktitle=	"Readings in Knowledge Representation and Reasoning",
 publisher=	"Morgan Kaufmann",
 address=	"San Mateo, California",
 year=		1985,


Another good paper to read would be:

 title=		"What's in a Concept: 
		 Structural Foundations for Semantic Networks",
 author=	"Brachman, Ronald J.",
 journal=	"International Journal of Man-Machine Studies",
 volume=	9,
 number=	2,
 month=		mar,
 year=		1977,


Peter F. Patel-Schneider
Bell Labs Research

Received on Sunday, 2 April 2006 21:06:27 UTC