Heurísticas para desambiguação incremental de nomes de autores em referências bibliográficas

Abstract

The ambiguity of author name in bibliographic references is one of the main problems affecting the quality of services offered by digital libraries. In recent years, numerous methods for automatic name disambiguation have been proposed based on different supervised and unsupervised approaches. However, just a few have been developed in order to allow the disambiguation at the time that citations are incorporated into the digital library. In a real situation, these solutions are potentially more efficient and practical, since they avoid the need to reprocess the entire repository whenever new citations are included in the database. This paper proposes a new incremental disambiguation method based on heuristics, able to automatically create and update a training set used to determine the author of each reference. This method was evaluated in different application scenarios and compared with several solutions found in the literature. In the experimental evaluation, our solution has achieved significant gains in all collections when compared with the best supervised and unsupervised baselines. We also performed experiments to demonstrate the practicability and efficiency of the method when used in a incremental way.