Provenance-Enabled Integration of Sensor Network Data
Authors
Pinheiro, Paulo
McGuinness, Deborah L.
ORCID
No Thumbnail Available
Other Contributors
Issue Date
2014-12-16
Keywords
E-Science Jefferson Project on Lake George
Degree
Terms of Use
Full Citation
Abstract
Ontology design in support of sensor networks requires deep scientific and technical knowledge. For instance, sensor network ontologies may define concepts rich enough to describe the selection of sensor locations that are driven by both scientific use cases, such as proximity to a physical entity, and technical requirements, such as wireless coverage. Our work in collaboration with IBM and the FUND for Lake George is set within the context of the Jefferson Project that aims to monitor, analyze, and understand observation and simulation data in support of decision making concerning the ecosystem of Lake George, NY. We have designed a Human-Aware Sensor Network Ontology (HASNet-O) that we believe is broadly reusable. We will highlight its contributions, describe its relationships to well used ontologies, and demonstrate it in action in the Jefferson Project. The ontology leverages best in class foundational ontologies including OBOE, VSTO, and PROV. One of the reasons HASNet-O is reusable is because it uses a provenance perspective to generalize what initially looked like sensor-specific terms that can be more accurately viewed as provenance terms. For example, we observe that the Activity concept from the PROV language, a W3C provenance standard, can be used to explain complete sensor lifecycles by describing how a network operates including how human interventions may interfere with the quality of the sensor network data, even for highly automated networks. We further observe that Observation activities, which are at the core of most sensor network ontologies, are not described in terms of deployments. In fact, Deployment activities are rarely documented in sensor data/metadata although Observation instances can only occur during the lifetime of Deployment instances. Our analysis of network ontologies in light of provenance knowledge includes a discussion about relationships. For instance, relationships between sensors and data are well-established provenance properties characterized in sensor network terms: the fact that a dataset was generated by a sensor may be described in terms of PROV’s wasDerivedFrom property between Activity and Entity instead of a less generic fromSensor relationship between Dataset and Sensor (in this case, it is assumed that Dataset is a subclass of PROV’s entity).
Description
Department
Publisher
Relationships
https://tw.rpi.edu/project/JeffersonProjectAtLakeGeorge