Análise de relacionamentos refinados do Tesauro AGROVOC: método de avaliação da qualidade em grandes tesauros

Abstract

Knowledge representation, based on Knowledge Organization Systems (KOS), experience restructuring, adaptation, and transformation to give support to the Semantic Web structure. Controlled, Thesaurus-style vocabularies follow this movement, as identified in a thesaurus reengineering approach. Based on the assumption that thesaurus is lacking in well-defined semantics and structural consistency, reengineering provides a refinement model for relationships between concepts. Inserted in the perspective of Architecture and Organization of Knowledge, this research has had for an object of research the refined relationships of the AGROVOC thesaurus. With the purpose of suggesting a method to evaluate the quality of refined relationships in thesauri, it strives to contribute to a semi-automatic evaluation of refined relationships in Simple Knowledge Organization Systems, so that it was possible to categorize existing relationships, analyze the precision of their instances and help the development of thesauri from the creation of a method to ensure quality. The methodological course, as to the choice of methods and procedures used for the development of the research, presents a formalization and systematization of the actions employed. For this, data extraction, analysis, and evaluation were done, aiming to find answers to the hypothesis that it is possible to develop a semi-automatic mechanism to verify the quality of refined relationships between concepts in a great thesaurus. In the investigation, it was possible to verify the main needs that a quality evaluation method in great thesauri demands. In the process of processing and preparing the data, it was possible to analyze the fundamental procedures for the proposal of an applicable method. The examination of the procedures performed resulted in a prototype developed on a data visualization platform. Finally, this study demonstrates that it is possible to use the method applied to elaborate a semi-automated system for evaluating the quality of relationships refined in large thesauri.