Semantically-Enabling the Web of Things: The W3C Semantic Sensor Network Ontology

grotesqueoperationInternet and Web Development

Oct 21, 2013 (4 years and 6 months ago)


Melbourne, Australia 6-10 Nov 2011
5th eResearch Australasia Conference
Semantically-Enabling the Web of Things: The W3C Semantic
Sensor Network Ontology
Kerry Taylor, Michael Compton, Laurent Lefort
CSIRO ICT Centre, Canberra, Australia,

The ecological and agricultural sciences, industrial processes, and consumer gadgets are increasingly relying on live data
streams generated by large numbers of heterogeneous sensors to deliver knowledge and services. All the traditional
problems of data management and data integration arise in this context of real time data, plus a few more. Semantic
technologies are being rapidly adopted for traditional data management and data integration problems, and there are
many international research projects now using semantic technologies for sensor network data management. The World
Wide Web Consortium (W3C) established an Incubator Group (SSN-XG) in March 2009 to develop ontologies for
describing sensors and methods for using those ontologies for annotation, especially in the context of the Open
Geospatial Consortiums (OGC) Sensor Web Enablement standards. The Group completed its work in June 2011 with
the publication of the final report, including the SSN OWL 2 ontology, use cases, extensive documentation and several
worked examples [1]. We present the ontology and some of the ways it is being used.
The SSN-XG comprised some 41 people from 16 organisations, with about 22 making active contributions to the
ontology itself over a period of about one year. Weekly international teleconferences, a single face-to-face meeting in
Washington DC, a lot of email correspondence and various e-collaboration tools were used. Use cases were proposed
and prioritised, existing sensor ontologies were reviewed [2], and then core classes and properties were developed first.
Group members volunteered for lead development of extensions to the core and proposals made by members were
presented, discussed, modified and voted on by the team before admission. The lead ontology editor was particularly
responsible for overall cohesion and maintenance of design principles set by the Group, as well as releasing the evolving
ontology versions. Some significant design principles were:
· Use local range restrictions on object properties instead of global domain and range property axioms;
· Source the names of classes from pre-existing standards and vocabularies where possible, and use ontology
annotation properties to document the source and any variation from that source;
· Support modularity and reusability by including only concepts relevant to sensors, identifying extension points
for inclusion of other ontologies in applications; and
· Align the ontology to an upper ontology (Dolce Ultralite (DUL) was chosen) to support that integration and to
further formalise the intention of SSN concepts.
The ontology design offers four identifiable perspectives, any of which could be the selected viewpoint in some
· Sensors, with a focus on what senses, how it senses, and what is sensed;
· Data, with a focus on observations and metadata;
· Systems, with a focus on systems of sensors; and
· Features, with a focus on physical features, properties of them, what can sense them, and what observations of
them are made.

Figure 1: Key concepts of the SSN ontology identified by module. Image: Raul Garcia Castro

Melbourne, Australia 6-10 Nov 2011
5th eResearch Australasia Conference
The concepts are informally grouped into modules for Systems, Deployment, Operating Restrictions, Processes, Devices,
Platforms, MeasuringCapability, and Data, around a central Core. Core concepts are sensors, sensing, sensing devices,
observations, properties of features that are sensed, and sensor inputs and outputs. Overall, the ontology defines 41
classes and 39 object properties, and imports many others from a module of DUL.

A sensor can do (implements) sensing: that is, a sensor is a physical object can enact a sensing method and thus observe
some property of a feature of interest. Sensors may be physical devices, computational methods, a laboratory setup with
a person following a method, or any other thing that can follow a sensing method to observe a property. Commonly, a
SensingDevice, that is both a sensor and a device and thus inherits properties from each, is the preferred way to describe
the artifacts we call sensors in sensor network systems. Our sensor corresponds directly with the concept of sensor in
the OGCs SensorML standard, but is a broader concept than a sensor in the OGCs O&M standard (with respect to
which to which our sensor is more like an observation procedure).

Our concept of observation was one of the hardest to develop: it is defined as a (DUL) situation in which a sensing
method has been used to estimate or calculate a value of a property of a feature of interest. Relations with sensing and
sensor describe what made the observation and how; relations also describe what was sensed, the result that is the output
of a sensor; and other obervational metadata such as the sampling time and a quality estimator.

The concepts MeasurementCapability and OperatingRange respectively collect together measurement properties
(accuracy, range, precision, etc) and the environmental conditions in which those properties hold, and the environmental
conditions and characteristics of a sensor's operating environment, such as a maintenance schedule or power needs. These
concepts are similar to respectively to the VIM [3] influence quantity and reference operating co ndition .

Examples of the ontologys application to several ontology
modelling problems are given in the Report[1]. These show
how to use various parts of the ontology, often in combination
with other ontologies or extended from the SSN ontology.
These examples are based around five scenarios:
· sensors deployed around a university campus for
location-based context inforrmation;
· smart consumer products which can sense, reason
and communicate;
· a high-end automatic weather station with multiple
sensing capabilities;
· agriculture meteorology where micro-climate sensing
is used to slelect and breed plants for food crops;
· 20,000 weather stations distributed throughout the
US that can be discovered and queried using
linked open data techniques.
For example, figure 2 shows the encoding of the 0.98g acceleration observed by a sensor which is embedded in a kitchen
knife in order to determine when it is cutting [4]. Typically this information would not be presented in XML as here
but this is the W3C standard encoding for OWL 2.
The members of the SSN-XG would be pleased to have their work widely used and the authors of this work would like to
hear from you if you do. Options for future development through the W3C and the OGC are being investigated.
1. Lefort L., Henson C., and Taylor K., (eds), The Semantic Sensor Network XG Final Report, W3C Incubator Group
Report 28 June 2011. Available from:
2. Compton, M., Henson, C., Neuhaus, H., Lefort, L., and Sheth, A., A survey of the semantic specification of sensors,
in Taylor et al (eds), 2
International Semantic Sensor Networks Workshop (SSN09) Washington DC, October 2009.
CEUR 522:p.17-32. Available from:

3. Working Group 2, International Vocabulary of MetrologyBasic and Gen eral Concepts and Associated Terms,
JCGM 200:2008, Joint Committee for Guides in Metrology 2008.
4. Sabou, M., Kantorovitch, J, Nikolov, A., Tokmakoff, A., X. Zhou, X., and Motta, E., Position Paper on Realizing
Smart Products: Challenges for Semantic Web Technologies. In Taylor et al (eds), 2
International Semantic Sensor
Networks Workshop (SSN09) Washington DC, October 2009. CEUR 522:p.135-147. Available from:

Figure 2: Acceleration of a kitchen knife

Melbourne, Australia 6-10 Nov 2011
5th eResearch Australasia Conference

Kerry Taylor has been a CSIRO computer scientist in CSIRO for over 15 years, working in areas broadly identified as
data management, most often with a focus on e-science. Kerry holds a BSc (Hons) in computer science from the
University of NSW and a PhD in computer science and information technology from the Australian National University,
where she is also currently an adjunct Associate Professor. Kerry leads the semantic data management research group in
the CSIRO ICT Centre, was a founding co-chair of the W3C Semantic Sensor Network Incubator Group, co-chairs the
series of Semantic Sensor Network Workshops held annually with the International Semantic Web Conference, and also
co-chairs the annual Australasian Ontology Workshop. In 2012 she will be chairing the Mobile and Sensor Web track of
the international Extending Semantic Web Conference (ESWC).

Michael Compton was awarded a BSc (Hons) in computer science and information technology from the Australian
National University and a PhD from the University of Cambridge, UK. His research work has focused on fundamental
aspects of logic, ontologies and other formal representation techniques, as well as applications in data translation, data
provenance modeling and sensor composition. Michael was the lead editor of the SSN ontology for the W3C Semantic
Sensor Network Incubator Group.
Laurent Lefort graduated in computer science (Engineering Degree) in 1983 from Ecole Nationale Supérieure
d'Informatique et de Mathématiques Appliquées de Grenoble (ENSIMAG), France and has been a researcher from the
CSIRO ICT Centre in Canberra since 2001. He is an ontologist in the team working on the application of semantic web
technologies to develop environmental sensor networks for agriculture meteorology, biodiversity, water and climate
change research. His current research interests include the design of ontologies, linked datasets, semantic mashups and
their use in data-intensive research service infrastructures. He has also served as a co-chair of W3C s Semantic Sensor
Network Incubator Group (XG) and as the W3C Australia Office manager.