Metodologia para construção de taxonomias para sistemas de recomendação: aprimorando a precisão com PLN e métricas de  audiência

• Ano de defesa: 2024
• Autor(a) principal: Mota, Mariana Silva
• Orientador(a): Souza, Osvaldo de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS::CIENCIA DA INFORMACAO
• Link de acesso: http://repositorio.ufc.br/handle/riufc/78710

This study aims to propose a methodology for building taxonomies using a two-phase methodological approach. The first phase, called Integrated UX Analysis, combines qualitative and quantitative methods to investigate user needs and identify market trends. The second phase, titled Taxonomy Methodology, focuses on developing the steps for creating taxonomies and defining their application scenarios in recommendation systems, with an emphasis on scalability and efficiency in such systems. The study seeks to propose a detailed methodology, outlining each step of the taxonomy-building process. Special attention is given to the influence of user experience on information structuring, with a focus on consumption patterns that directly impact e-commerce. The study also covers specific objectives such as understanding the user context through user experience (UX) techniques and the analysis of terms extracted via natural language processing (NLP), assessing the performance of these terms in recommendation systems. This study employs various tools, including surveys, proprietary databases, quantitative and qualitative analysis, as well as market studies, allowing for a comprehensive analysis of user contexts and the subsequent information modeling for taxonomy creation. The results highlight the central importance of the user in the development of products and services, demonstrating that a well-structured taxonomy not only facilitates categorization and machine learning but also enables efficient information consumption. The proposed methodology proves to be easy to implement while offering great potential in terms of flexibility for different scenarios and precision in the taxonomy-building steps, emphasizing the relevance of Information Science knowledge, such as taxonomy, in advancing technological solutions and optimizing recommendation systems. Finally, the study contributes to the validation of the proposed methodology, expanding the understanding of the interactions between Information Science and technology, while also opening new opportunities for future research in this field.