A extração de relações semânticas como suporte para a construção de estruturas conceituais: um estudo de caso na Intensificação Agropecuária

Abstract

Knowledge Organization (KO), in the context of Library and Information Science (LIS), aims to systematize and structure information to facilitate access, understanding, and the efficient use of knowledge in different contexts. KO seeks to categorize, describe, and relate concepts, documents, and data, providing a structure that assists in the precise and relevant retrieval of information, whether in libraries, information systems, or digital environments. This research aims to apply the Semantizar tool to present indications of relationships between concepts in the context of the book "Agrarian Evolution and Population Pressure" by Ester Boserup. The automation of the process of identifying and making explicit the connections between concepts and entities, promoted by relation extraction tools such as Semantizar, enhances KO’s capabilities by allowing large volumes of data to be semantically organized, promoting scalability, consistency, and continuous updating of conceptual structures. The methodology is classified as applied, given its intention to solve practical problems, and exploratory, as it investigates the tool. It employs a mixed qualitative and quantitative approach. In the mixed approach, the concept list guides the identification of relevant terms, while the gold standard structure allows for a comparison between automated results and the manual model. To examine the tool's application, a case study was used as a technical procedure. For this purpose, an application protocol was built based on a literature review. The results indicate 31 established relationships, achieving precision, recall, and F1-score of 0.21, 0.37, and 0.1339, respectively. Although below the levels found in comparative studies, the tool demonstrates potential by suggesting relationships absent from the reference standard, thus expanding the conceptual structure. It is concluded that, although promising, the tool requires adjustments to improve its accuracy, highlighting the importance of advances in semantic relation extraction.