A4: um ambiente de apoio à aquisição automática de conhecimento.

Abstract

Concept learning methods have been strongly used for knowledge acquisition in the knowledge based systems. However, these methods present some problems due to constraints in the real world modeling, the inductive algorithm diversity and the form as their results are presented. Besides, the domain modeling for knowledge acquisition inductive methods has not been properly approached. These methods have approached only the generalization process. However their success depend on the quality of the domain modeling. In this work we propose the A4 (Knowledge Acquisition Environment Support) that helps the inductive knowledge acquisition process, modeling the domain into examples and background knowledge and it treats the algorithms results. A4 is based on the integration of interviews and machine learning methods in order to get the advantages of both. A4 architecture is object oriented. The three basic class are: domain modeling, inductive algorithm and output treatment. The use of object oriented paradigm added new functionalities to the environment such as procedure reutilization. Mainly, the domain modeling class is studied. This class helps the real word modeling into examples and background knowledge. This modeling increases the automatization degree of the learning process which produces faster and more powerful results. The A4 has been implemented in C++ in the SPARC/SUN workstation.