An automatic approach for duplicate bibliographic metadata identification using classification

Abstract

References are the main descriptive metadata used by digital libraries of scientific articles. These references can be represented by several formats and styles. Although considerable content variations can also occur in some metadata fields such as title, author names and publication venue. Duplicate records influence the quality of digital library services once they need to be appropriately identified and treated. This paper presents an approach to identifying duplicated bibliographic metadata. We extend our previous work so that instead of setting thresholds based on the scores returned by similarity functions, we use the scores to train classification algorithms which automatically identify duplicated references. The experiments show that the classifiers increases up to 11% the quality of results when compared to our unsupervised heuristic-based approach.