LIP6 1998/018
- Thesis
Connaissances taxinomiques: Représentation de taxinomies comportant des exceptions et construction d'hypermédias à base de connaissances taxinomiques - C. Faron
- 187 pages - 05/05/1998- document en - http://www.lip6.fr/lip6/reports/1998/lip6.1998.018.ps.gz - 641 Ko
- Contact : Catherine.Faron (at) nulllip6.fr
- Ancien Thème : APA
- Keywords : Artificial Intelligence, Hypermedias, Knowledge Representation, Knowledge Acquisition, Conceptual Graphs, Description Logics, Exceptions, Taxonomies, Classification
- Publisher : Valerie.Mangin (at) nulllip6.fr
A taxonomic structure is inherent to many domains.: objects are clustered into classes hierarchically organized into taxonomies. The problem addressed by these domains is to memorize this taxonomic knowledge, that is descriptions of classes and their taxonomic organization. In artificial intelligence, one field is precisely the elicitation and representation of knowledge. To acquire taxonomic knowledge, given the domain taxonomy, we elicit descriptions of its concepts whose order relations among them account for the taxonomy structure. To represent this knowledge, we have chosen the Conceptual Graphs model that suits particularly well these domains whose concepts require structured descriptions, as it is the case of natural taxonomies.
A problem specific to natural taxonomies is the representation of exceptions to properties inheritance between classes. To do this, we propose an extension of the Conceptual Graphs model consisting in the introduction of defaults and exceptions in graphs defining concepts. The so called definitional approach we here adopt is opposed to the inferential one consisting in default inheritance.: the classical principles of classification are not modified. We show that some approaches adopted with Description Logics can be extended and some results stated in these formalisms can be adapted to Conceptual Graphs.
The elicited taxonomic knowledge is dedicated to the organization of domain information into an hypermedia network. We propose a method for the construction of taxonomic knowledge based hypermedia, based on the method we use for knowledge acquisition and on a typage of the hypermedia elements that reflects the structure of the knowledge base represented with concetual graphs. Such a knowledge base hypermedia offers the user a conceptual navigation and access procedures based on the inference mechanisms associated to the knowledge base.
A problem specific to natural taxonomies is the representation of exceptions to properties inheritance between classes. To do this, we propose an extension of the Conceptual Graphs model consisting in the introduction of defaults and exceptions in graphs defining concepts. The so called definitional approach we here adopt is opposed to the inferential one consisting in default inheritance.: the classical principles of classification are not modified. We show that some approaches adopted with Description Logics can be extended and some results stated in these formalisms can be adapted to Conceptual Graphs.
The elicited taxonomic knowledge is dedicated to the organization of domain information into an hypermedia network. We propose a method for the construction of taxonomic knowledge based hypermedia, based on the method we use for knowledge acquisition and on a typage of the hypermedia elements that reflects the structure of the knowledge base represented with concetual graphs. Such a knowledge base hypermedia offers the user a conceptual navigation and access procedures based on the inference mechanisms associated to the knowledge base.