Sistema de recomendação hí­drido para bibliotecas digitais que suportam o protocolo OAI PMH

Abstract

The growth of Web technologies has benefited researchers and the academic community by supporting the access of electronic publications as soon as they have been finished and published. In this context, Digital Libraries emerges as complex information systems which are essential for disseminating and preserving data, information and knowledge. However, due to the high amount of content available on the Web, specially in Digital Libraries, users face many correlated options, what result in the phenomenon known as information overload. Aiming to decrease or even eliminate these diffculties, recommender systems for Digital Libraries have been proposed and developed. This work presents a personalized recommender system which presents alternative ways to achieve better query results. For this, the main existing approaches of automatic recommendation have been studied in order to identify extension points and points to be improved. The proposed recommender system follows a hybrid approach which combines filtering techniques, content-based recommendation and collaborative recommendation. A hybrid recommendation engine has been proposed, which uses standard technologies for content description (Dublin Core), for communication with Digital Libraries (OAI-PMH Protocol ), as well as the user profile extracted from the curriculum vitae Lattes. The proposed solution has been evaluated in the context of the CiteSeer database, which contains papers and articles in the Dublin Core format. The preliminary results has showed an improvement in the quality of recommendation, thus presenting a better precision and coverage, when compared with existing approaches based either on content-based recommendation or on collaborative recommendation