Uma abordagem para identificar a similaridade entre perfis de pesquisadores com vistas a recomendação

• Ano de defesa: 2018
• Autor(a) principal: Rovadosky, Diogo Nelson
• Orientador(a): Cervi, Cristiano Roberto
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade de Passo Fundo
• Programa de Pós-Graduação: Programa de Pós-Graduação em Computação Aplicada
• Departamento: Instituto de Ciências Exatas e Geociências – ICEG
• País: Brasil
• Palavras-chave em Português: Software; Pesquisadores
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://tede.upf.br/jspui/handle/tede/1477

The production of knowledge by humanity increases with each passing day, generating a huge impact in the process of discovering the new knowledge produced. The internet has made it easier to share information, but its size now tends to make this whole thing difficult. In this scenario, scientific production has been gaining ground, raising the need for a process of qualification and deeper analysis of available data. Database with this information appears to facilitate access to the latest searches and provide relevant information. However, there is much work to be done to make this information increasingly relevant to researchers and research promotion agencies. That is why in this context, recommendation systems become promising for their ability to help resolve information overload by identifying, in a contextspecific, context-sensitive information, and customizing content based on similar profiles stands out, both for recommending similar articles and for suggesting new partners for research. To do so, we used the data from the researcher's curriculum to build his profile, which expresses his preferences through the history of scientific life, indicating his academic future, his preferences regarding research and points future connections with other researchers. Thus, this paper aims to present an approach to identify the similarity between profiles of researchers, which allows the generation of recommendations based on their profiles. In addition, we tried to structure a profile model of researchers to identify similarities between the profiles, proposing a metric to calculate such similarities, with the final objective of using the profile model and the similarity metric next to the recommendation mechanism. And, create a tool that uses the profile model, the similarity metric with the recommendation engine. For the development of the approach, we used personalization concepts and recommendation systems. Through experiments carried out with real data from researchers from 8 areas obtained from the Lattes Platform, it was observed that the proposed approach tends to help the discovery of useful content to the researcher through recommendati ons. Experiments also demonstrate that the proposed approach has a good coverage of recommendations. In the same way that, through calculations with the terms mined in the curriculum, it was possible to apprehend the preferences and identify changes in them, when analyzing the curriculum of the researcher along temporal aspects. Not only do similarity indicators point to similarities between profiles, but they also help to generate recommendations.