Authors:
Heba Mohamed
1
;
2
;
Said Fathalla
1
;
2
;
Jens Lehmann
3
;
2
and
Hajira Jabeen
4
Affiliations:
1
Faculty of Science, University of Alexandria, Alexandria, Egypt
;
2
Smart Data Analytics (SDA), University of Bonn, Bonn, Germany
;
3
Fraunhofer IAIS, Dresden, Germany
;
4
Cluster of Excellence on Plant Sciences (CEPLAS), University of Cologne, Cologne, Germany
Keyword(s):
In-memory Computing, Distributed Processing, Hadoop Streaming, Ontology Parsing, SANSA Framework, Large-scale Datasets.
Abstract:
Ontologies are widely used in many diverse disciplines, including but not limited to biology, geology, medicine, geography and scholarly communications. In order to understand the axiomatic structure of the ontologies in OWL/XML syntax, an OWL/XML parser is needed. Several research efforts offer such parsers; however, these parsers usually show severe limitations as the dataset size increases beyond a single machine’s capabilities. To meet increasing data requirements, we present a novel approach, i.e., DistOWL, for parsing large-scale OWL/XML datasets in a cost-effective and scalable manner. DistOWL is implemented using an in-memory and distributed framework, i.e., Apache Spark. While the application of the parser is rather generic, two use cases are presented for the usage of DistOWL. The Lehigh University Benchmark (LUBM) has been used for the evaluation of DistOWL. The preliminary results show that DistOWL provides a linear scale-up compared to prior centralized approaches.