ホァキン デルガド, ジュリ ティヘリノ
Knowledge Representation and Acquisition
for 3-D shape Ontologies
Abstract:In order to build more intuitive virtual environments, there is a need for the computer to understand the meaning of virtual objects in a way that reflects common sense knowledge of such objects. This report introduces several mechanisms for building a 3-D shape ontology that eventually would add such common sense knowledge to the What You Say is What You See (WYSWYS) framework [1]. Principally, this can be achieved through ontologies of primitive 3-D shapes and of combination of such shapes into more complex shapes. The representation used for expressing 3-D shapes is accomplished through a two-level ontology approach, that is a knowledge level and a symbol level, where the former is a representation closest that of humans while the later is a high level computer representation in the form of implicit parametric functions. Under the same approach, this report proposes a more detailed ontology hierarchy, mainly concentrating on the evaluation of formal methods for the edition and construction of the symbol level ontology using Ontolingua [2], and in knowledge acquisition through knowledge reuse and machine learning techniques, for building the knowledge level ontology. It presents a modification of the case based reasoning (CBR) algorithm PROTOS [3] for this last purpose. To the end, this report also gives guidelines for future implementation and further research topics.