ARQUITETURA BASEADA EM ONTOLOGIAS DE UM AGENTE RBC

Abstract

Case-Based Reasoning (CBR) is a problem-solving paradigm where it is possible to use knowledge from past experiences to solve new situations. The CBR agent approach that combines agent autonomy with the problem-solving model of CBR has been proven adequate for the development of complex systems. This paper proposes the architecture of a CBR agent whose main differential is the use of ontologies for representing the case base along with all the mechanisms that make up a CBR system. The proposed architecture besides promoting the reuse of the case ontology, unifies the software agent approach with CBR, a typical paradigm of human reasoning. All the CBR mechanisms are present in the proposed architecture: case representation, similarity analysis for cases retrieval, adaptation and cases learning, where the last two mechanisms are still being specified. The architecture was evaluated in the Brazilian Family Law legal domain. For that, a targeted ontology for the representation CBR cases of this area was created. The results obtained in the tests showed good effectiveness in retrieving similar cases and showing the feasibility of the architecture using the semantic model of similarity for retrieval of CBR cases.