1. Les formalismes de représentation de connaissances
Une manière courante de représenter des connaissances externes à un ordinateur ou à un être humain est sous forme de langage écrit. Par exemple, certains faits et relations représentés en français (écrit) sont :
omar est grand //expression d’un fait simple grand est une proprité de omar.
Amine aime son fils./ expression d’une relation
Amine a appris à utiliser la récursivité pour manipuler des listes chaînées dans plusieurs langages de programmation.// plus complexe relation entre amine et concept abstrait (langage de programmation)
Les formalismes de représentation de connaissances
We do not know how knowledge is represented in the human brain,
but we can define some characteristics that will be important in representing problem-solving knowledge. Clearly, we will have to represent
facts, such as "an Alfa-Romeo is a car" or "Peter is a man." Furthermore, we must be able to represent relationships between these facts,
such as "Peter owns an AHa-Romeo." In reality, relationships can be
very complex, such as the management structure of a large organization or the structure of a huge organic molecule. Facts and relationships are obviously important, but they have limited value in themselves. It is the sort of knowledge we find in the data division of a COBOL program. It is not capable of solving problems. To solve problems we need knowledge that acts upon the facts and relationships and creates new ones. Such knowledge is equivalent to the procedure division of a COBOL program.
The procedural knowledge of a FORTRAN or COBOL program can be
represented in a variety of constructs.
Langage natural
Une manière courante de représenter des connaissances externes à un ordinateur ou à un être humain est sous forme de langage écrit. Par exemple, certains faits et relations représentés en français (écrit) sont :
omar est grand //expression d’un fait simple grand est une proprité de omar.
Amine aime son fils./ expression d’une relation
Amine a appris à utiliser la récursivité pour manipuler des listes chaînées dans plusieurs langages de programmation.// plus complexe relation entre amine et concept abstrait (langage de programmation)
Autres formalisme
Triplets <objet, attribut, valeur>
Règles
Réseaux sémantiques
Frames
Logique
Types de logiques
Langages
Logique propositionnelle Faits vrai/faux/inconnu
Logique des prédicats Faits, objets, relations vrai/faux/ inconnu
Logique temporelle Faits, objets, relations, temps vrai/faux/inconnu
Théorie des probabilités Faits Degré de croyance 0 .. 1
Logique floue Degrés de vérité Degré de croyance 0 .. 1
2. Knowledge Representation in Natural Language
Humans usually use natural language (English, Spanish, Chinese, etc.) to represent
knowledge, so why not use that to represent knowledge in our AI systems?
Advantages of Natural Language
1. It is extremely expressive – we can express virtually everything in natural language
(real world situations, pictures, symbols, ideas, emotions, reasoning, …).
2. Most humans use it most of the time as their knowledge representation of choice
(how many text books are not written in natural language?).
Disadvantages
1. Both the syntax and semantics are very complex and not fully understood.
2. There is little uniformity in the structure of sentences.
3. It is often ambiguous – in fact, it is usually ambiguous
3. Databases as a Knowledge Representation
Traditional database systems are clearly very powerful, but for AI systems they are rather limited. The important issues are:
Advantages
1. Databases are well suited to efficiently representing and processing large amounts
of data (and derivation from a database is virtually independent of its size).
2. We can build on traditional database systems to process more complex and more
powerful representational devices (e.g. frames).
Disadvantages
1. Only simple aspects of the problem domain can be accommodated.
2. We can represent entities, and relationships between entities, but not much more.
3. Reasoning is very simple – basically the only reasoning possible is simple lookup,
and we usually need more sophisticated processing than that.
4. Triplets
Sert à représenter un fait : un énoncé vrai ou faux.