DESiRe : uma abordagem dinâmica para recomendação de resultados de busca em atividades de pesquisa exploratória

• Ano de defesa: 2015
• Autor(a) principal: Nanni, Lucas Pupulin
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual de Maringá; Brasil; Departamento de Informática; Programa de Pós-Graduação em Ciência da Computação; UEM; Maringá, PR; Centro de Tecnologia
• Palavras-chave em Português: Recomendação dinâmica de resultados de busca; Recuperação da informação; Pesquisa exploratória; Busca personalizada; Recuperação interativa de informação; Brazil.; Interactive information retrieval; Personalized search; Exploratory Search; Dynamic search results recommendation; Ciências Exatas e da Terra; Ciência da Computação
• Link de acesso: http://repositorio.uem.br:8080/jspui/handle/1/2505

The difficulty to retrieve relevant and quality information on the Web has worsened as the complexity and availability of information resources increase. Much of such difficulty stems from the deficiency of search engines to identify the informational context of the search performed by the user. This problem increases when the user is engaged in complex search activities such as Exploratory Search. To overcome this deficiency, several studies have proposed tools and techniques related to adaptive retrieval and personalized search concepts in order to capture the different users' information needs and provide them with relevant results considering their individual interests. However, works done so far restricted themselves only to the results page to display the customization efforts, failing to explore the possibility of recommending results while the user navigates through the search space. Thus, we propose Desire, a dynamic approach for recommending search results able to present relevant results while the user navigates through the retrieved search space. The approach is based on a model of the user's search interest to re-rank the retrieved documents, allowing to recommend a set of relevant results to the user during navigation. To make dynamic the process of recommendation, the model is built as the user navigates the retrieved documents and informs his interest about them. The results obtained by simulations have shown that our approach provided an average improvement of 88% in the quality of the ranking provided by Google. In addition, the recommendation process was equally effective, providing high quality recommendations with a relatively high number of unseen results. Preliminary evaluations have also shown that the approachis able to handle suspect feedbacks and to propagate the model of interest among the several activities that can comprise the same search task