Knowledge base

Knowledge Base

Expert knowledge about the best localization conditions (``context'') for the solution of a problem is expressed via production rules.

Each rule (``knowledge unit'') is given an "Importance Degree", noted ID_rule , depending on the relative necessity of the corresponding context for the solution :


totally necessary	+1.0	totally inadequate	-1.0
highly necessary	+0.8	highly inadequate	-0.8
very necessary	+0.6	very inadequate	-0.6
quite necessary	+0.4	quite inadequate	-0.4
somehow necessary	+0.2	somehow inadequate	-0.2
neutral	0.0	-	-

Note that this importance degree may be obtained from statistical human observations . For previous similar problems, if the solution has been found in a given context, the importance degree associated represents a sort of frequency degree of appearance of this context.

For example :

"Being near a road is quite necessary" will be translated in the following production rule :

IF Solution THEN ( near road ) ( =0.7)

The expert knowledge will be translated in a set of such production rules. Each conclusion of a rule describes a particular context whose relative likelihood can be computed on the image or the G.I.S. by adequate procedures.

Note that conclusion of a rule may consist in logical combinations of elementary premises. The possible forms of rules are as follows :

and B

or B) and C

For example :

"Principally south slopes and elevation from 800 to 1000 m." will be translated in the following rule :

IF solution THEN (South slopes) and (800 <elevation<1000) ( = 0.8)

This knowledge base may be built by means of an interface (rule compiler) where the expert expresses his knowledge in natural language but with predefined keywords.

IRIT-UPS