Skip to Main content Skip to Navigation
Theses

Raisonnement classificatoire dans une représentation à objets multi-points de vue

Abstract : In a taxonomy, knowledge is structured in categories grouping similar objects. These categories are organized in a specialization graph, going from general to more specific classes. Taxonomies allow for classification reasoning. To reason by classification is to find the more accurate class for an instance and to infer knowledge from this localization. Taxonomies developed in areas as different as botanic and mineralogy show the interest of this approach. Our work concerns classification reasoning and the taxonomic representation supporting it. We chose object oriented representation for it provides the elements required for modeling taxonomies. Moreover, it is a natural classification framework. However, these models present two problems: on one hand, different object families, such as "cars" and "persons", are represented in a single big taxonomy. On the other hand, although object features correspond to different aspects or viewpoints, these viewpoints are not explicitly represented in the models. We propose an object oriented multi-perspectives representation system, TROPES. In this system, each concept or object family has an independent taxonomic structure. A concept may be observed from different viewpoints, each one focusing on a set of features and a structure of categories. Viewpoints may be linked by bridges. Furthermore, the viewpoint notion of TROPES rules out multiple
Personal Toolbar Folderinheritance problems. TROPES is provided a multi-perspectives instance classification algorithm which takes advantage of the model originalities. Inside a concept, the classification takes place on one or more viewpoints. Bridges are used to fasten the classification.
Document type :
Theses
Complete list of metadata

https://tel.archives-ouvertes.fr/tel-00005133
Contributor : Thèses Imag <>
Submitted on : Thursday, February 26, 2004 - 4:18:14 PM
Last modification on : Friday, November 6, 2020 - 3:53:59 AM
Long-term archiving on: : Friday, September 14, 2012 - 1:11:11 PM

Identifiers

  • HAL Id : tel-00005133, version 1

Collections

UJF | CNRS | UGA

Citation

Olga Marino Drews. Raisonnement classificatoire dans une représentation à objets multi-points de vue. Interface homme-machine [cs.HC]. Université Joseph-Fourier - Grenoble I, 1993. Français. ⟨tel-00005133⟩

Share

Metrics

Record views

455

Files downloads

1286