Uma estratégia para remoção de ambiguidades na identificação de autoria de objetos bibliográficos

Abstract

The problem of informational overload generated by the success of the Web has led to the emergence of services that congregate information in specific contexts, known as digital libraries. Digital libraries combine digital information and metadata that frequently are collected from diverse sources. The lack of standardization of metadata deriving from these sources brings as consequence the ambiguity in determined fields. In this dissertation we present a strategy for the disambiguation in fields referring to names of authors in digital libraries. Our strategy uses Information Retrieval techniques associated to a clustering algorithm that allows the creation of unified indexes. We demonstrate the effectiveness of our strategy through a set of experiments conducted on two test collections derived from the Biblioteca Digital Brasileira de Computação (BDBComp) and the Digital Bibliography of Library Project (DBLP). For the BDBComp collection, the average between the quality of the generated groups and its fragmentation was over the mark of 95%, and for the collection of the DBLP, this average was over 65%.