Automatic RDF-ization of big data semi-structured datasets

Abstract

Linked data adoption continues to grow in many fields at a considerable pace. However, some of the most important datasets usually remain underexploited because of two main reasons: the huge volume of the datasets and the lack of methods for automatic conversion to RDF. This paper presents an automatic approach to tackle these problems by leveraging recent Big Data tools and a program for automatic conversion from a relational model to RDF. Overall, the process can be summarized in three steps: 1) bulk transfer of data from different sources to Hive/HDFS; 2) transformation of data on Hive to RDF using D2RQ; and 3) storing the resulting RDF in CumulusRDF. By using these Big Data tools, the platform will cope with the handling of big amounts of data available in different sources, which can include structured or semi-structured data. Moreover, since the RDF data are stored in CumulusRDF in the final step, users or applications can consume the resulting data by means of web services or SPARQL queries. Finally, an evaluation in the hydro-meteorological domain demonstrates the soundness of our approach.