Conference heron::euro_swas_ai

188.0. "Fuzzy database and query ... your opinion?" by KETJE::VANDEVYVER (Luc Van de Vyver 856-7577) Tue Apr 03 1990 18:27

An IT-department of a Belgian university did some testing/study/experiment
with what they call a 'fuzzy database and query system'

A fuzzy database contains besides exact data, some approximate data.
Like : persons skin color : pale;
       persons hair color : dark brown
       persons length     : tall
       beard/moustache    : probably yes
       eye color          : unknown
       .....

The queries to be executed on this kind of data may look like :
	- All RATHER TALL persons with DARK hair and probably a beard but no
          moustache .....
Thus using similar approximates. If DARK hair is selected, DARK brown, black,
purple, ... should come out.

To define if the criteria is more or less met, the following model is used :

  |
  |      ______
  |     /|    |\
  |    / |    | \
  |   /  |    |  \
  |  /   |    |   \
  | /    |    |    \ 
  +---------------------
    0%    100%    0%

so, in the above example, purple hair may score 30%.


They've created an example using LISP on Appollo-domain for storing information
about criminals. 
The database and the query system are written in LISP (with low performance).

The idea is to develop a generic-fuzzy database, based on existing products like
RDB (for easy access to non-fuzzy data), and a shell on top of it to allow the
fuzzy queries. The shel could be developed using an expert system shell, lisp,
ops, or whatever.

The university department is currently looking for a partner for such development
and marketing afterwards.

What do you think about such a system?
Is it feasable? Does one (useable) exists already?
Any opinion is welcome.

Thanks
Luc

    
    Is Luc Steels involved here ?
    
    -Malcolm
    

Luc Steels is not involved in this 'project'.

The university involved is RUG : Rijks Universiteit Gent

Luc