Múltiplos sistemas de raciocínio baseado em casos para autenticação de pinturas

Abstract

The development of Case-Based Reasoning (CBR) systems is suitable when the solution for a problem can be found from the retrieval and reuse of past experiences of problem solving. In this context, the main goal of this Undergraduation Dissertation is to develop CBR techniques to support the process of date authentication of historical paintings. In particular, this dissertation exploits the automatic or semi-automatic acquisition of data and knowledge coming from experiences of problem solving, which are represented as "cases" in a web-based CBR system called AuthenticPainting. Furthermore, this work describes how to develop the analysis of similarities between different types of typical attributes found in painting authentication reports, which in this dissertation are historical pigments and elemental compositions of paintings. In general, this dissertation aims to combine and test the results that can be obtained when multiple mechanisms of case retrieve, or multiple CBR systems, based on these attributes are used on the solution of new problems. The results obtained allow one to conclude that the use of multiple CBR systems, along with the combination method which was tested with these systems, is relevant to assist the authentication of painting dates when one of the involved CBR systems is able to produce results that may be confirmed by the solution presented by the other CBR system. Keywords: