Semantics for big data applications in the smart dairy farming domain

In the Dutch SmartDairyFarming project, better decision support for the dairy farmer on daily questions around feeding, insemination, calving and milk production processes is made possible using a variety of big data sources containing static and dynamic sensor data of individual cows. This paper is concerned with the inherent semantic interoperability problem between the information in these data sources. Semantic alignment is achieved using ontologies and linked data mechanisms on a large amount of sensor data, such as grazing activity, feed intake,weight, temperature and milk production of individual cows at 7 dairy farms in The Netherlands. A Common Dairy Ontology (CDO) and a specific measurement ontology have been developed and used to transform the sensor data into 310GB of RDF triples, made accessible via a SPARQL interface on a triplestore. A few example applications have been developed to show how the CDO can be used for decision support and historic analysis. The performance of our linked data semantic solution is acceptable for analysis queries on large sets of data. Without optimization of queries the time for answering queries ranged from a few seconds to a couple of minutes