Re: The stupidest design I ever saw

From: Alexandr Savinov <spam_at_conceptoriented.com>
Date: Tue, 11 Apr 2006 15:21:50 +0200
Message-ID: <443bad6e$1@news.fhg.de>

Mikito Harakiri schrieb:
> Neo wrote:

>>>>> Mikito: Cartoon is a mammal ...
>>>> Neo: Cartoon is not a mammal.
>>> Mikito: Look into the table on figure 1
>>> which is the basis for the lattice on figure 2.
>>> If you don't want cartoon to be classified as mammal,
>>> which squares should be unchecked?
>> First there is no systematic method that allows one to create the
>> hierarchy (fig 2) on the basis of the table (fig1) (even after ignoring
>> the top most and bottom most nodes which represent Universal and NULL
>> respectively). If you know what that method is, please state it so that
>> anyone can verify it.

>
> The lattice nodes on figure 2 correspond to "cylindric sets" on figure
> 1. Each cylindric set is a cartesian product of some set of objects
> with some set of attributes. For example the unnamed node at the right
> of the "dog" corresponds to a set
> {Socks,G Bobby}x{real,mammal}. Therefore, the formal method of
> transfoming figure 1 into the figure 2 is as follows:
> 1. Identify all the cylindric sets

It is an exponentially difficult procedure. There exist many algorithms for finding this kind of sets including approximate ones. It is a rather general problem which is formulated and studied in other domains. Yet, it belongs more to data analysis rather than data representation and manipulation.

> 2. Connect nodes with the partial order relation

The number of identified nodes is also exponential. Only a tiny portion of them is semantically meaningful and formally important. And many of them will be very similar.

> The partial order is defined as follows. If the set of attributes A1 is
> a subset of A2
> and if the set of objects O2 is a subset of objects O1, then A1xO1 and
> A2xO2 are ordered. In the example {Socks} is subset of {Socks,G Bobby}
> and {real,mammal} is subset of {cat, real, mammal}. Therefore, the
> nodes {Socks,G Bobby}x{real,mammal} and {Socks}x{cat, real, mammal} are
> ordered -- in fact there is a lattice edge between this unnamed node
> and the object "Socks".
>

>> Second, you can't unclassify cartoon as a mammal in the table because
>> it isn't being classified as an mammal. Both cartoon and mammal are
>> classifications listed across the top. And as the author states "The
>> elements on the left side are formal objects; the elements at the top
>> are formal attributes [aka classes]; and the relation between them
>> [objects and classes] is represented by the crosses." Please state THE
>> square/cell (and there should be only one, if any at all, otherwise why
>> the redundancy) to uncheck which unclassifies cartoon as a mammal
>> without also unclassifying Garfield and Snoopy as cartoons? A portion
>> of the table shown below:
>>
>> ______ cartoon  real   dog   cat   mammal
>> Garfield __X___  ___   ___   _X_   __X____
>> Snoopy __X___  ___  _X__  ___   __X____

>
> This is not big deal. Any classification system has to obey
> transitivity rule: if class A is a subclass of B and if B is subclass C
> then A has to be a subclass of C. Garfield, which is a cartoon is
> defined to be a cat. Cat is defined to be a mammal. Therefore, Garfield
> is a mammal, even though it is a cartoon. If you want a more realistic
> example where cartoon is not a subclass of mammal, or insist that these
> categories don't intersect at all, then you have to recategorize these
> cartoon characters as "cartoon cat" and "cartoon dog" which are not
> mammals.
>

-- 
http://conceptoriented.com

Received on Tue Apr 11 2006 - 08:21:50 CDT