Alinhamento de dados conectados a partir de conceitos primários e secundários

Abstract

Recently linked data has been the most prominent way to open data in several countries. This way it uses standards to describe this data, promoting its interoperability, its reuse and its integration. However, integrating information between different datasets is a hindrance to their development, especially if such integration consists of matching a particular real-world entity in distinct datasets. In this context, this work proposes an approach to assist in the identification of corresponding instances. For this, it is based on the conceptual modeling of the data, allowing the relationships between the concepts to be used to discover new correspondences between the data. To evaluate the effectiveness of the proposal, a case study and an experiment were carried out. In the case study, the proposal was used to find the correspondence of researchers and publications in four datasets (Lattes, RBIE, SBIE and WIE). These correspondences was used to answer a set with more than thirty questions provided by the informatics community in Education. In the experiment, the proposal was used in two scenarios (C1 and C2) and compared to other approaches through precision, recall and f-measure metrics. According to the presented results, the proposal ranked first and second place in scenarios C1 and C2 respectively, even though it did not use specific computations for the datasets, allowing its use in other contexts with the least effort.