ios press sense2web: a linked data platform for semantic … · undeﬁned 0 (0) 1 1 ios press...

Undefined 0 (0) 1 1IOS Press

Sense2Web: A Linked Data Platform forSemantic Sensor NetworksPayam Barnaghi a,∗, Frieder Ganz a, Hamidreza Abangar a, Mirko Presser b, and Klaus Moessner a

a Centre for Communication Systems Research, University of Surrey, Guildford, GU2 7XH, United Kingdom.E-mail: {p.barnaghi, f.ganz, h.abangar, k.moessner}@surrey.ac.ukb The Alexandra Institute Åbogade 34, 8200 Århus N, Denmark.E-mail: {mirko.presser}@alexandra.dk

Abstract. The current advancements in sensor networks and being able to manufacture low cost and energy efficient hardwarefor sensors has lead to a potential interest in integrating physical world data into the Web. The sensor networks range from tinywireless artefacts to mobile devices (with many sensors) to large scale systems (in smart cities and environmental monitoringnetworks). Applications and users are typically interested in querying various events and requesting measurement and observationdata from the physical world. Integration of data from physical objects, using embedded processors and sensors, into the Web isgiving rise to the emergence of a new generation of systems. The incorporation of physical object data with the Web data, usinginformation processing and knowledge engineering methods, enables the construction of "‘intelligent and interconnected things"’and "‘smart environments"’. The current research mostly focuses on the communication and networking aspects between suchdevices. There is, however, relatively less effort concentrated on creating dynamic infrastructures to support integration of thedata into the Web and provide unified access to such data on service and application levels. Utilising semantic Web technologiesin the sensor networks results in a new concept, sometimes referred to as Semantic Sensor Networks. This paper describes aplatform to publish Semantic Sensor Network data and to link the data to existing resources on the Web. The linked SemanticSensor Web data platform, called Sense2Web, supports the publication of extensible and interoperable resource (i.e. sensornetwork and service resources) descriptions and observation and measurement data in the form of linked data. Sense2Web alsosupports the association of different sensor data to resources described on the Web of Data. In this paper we focus on publishinglinked data to describe sensor data and sensor network resources descriptions and link them to other existing resources on theWeb.

Keywords:, Linked data, Ontology, Sensor, Semantic Sensor Web, Sensor Networks, Sense2Web

1. Introduction

The information collected from the physical world,in combination with the existing resources and ser-vices on the Web, facilitates enhanced methods of ob-taining business intelligence, enabling the constructionof new types of front-end applications and services,and could revolutionise the way organisations and peo-ple use Internet services and applications in their dailyactivities. There are currently a number of projects fo-cused on developing large-scale sensor networks in-

*Corresponding author. E-mail: [email protected]

tegrated into the Internet such as SENSEI1, IOT-A2,SensorWeb3, and also there are existing work on cre-ating service layers and network structures for sensorand actuator networks such as [1], [2], [3], the OpenGeospatial Consortium (OGC)4 and the Sensor WebEnablement (SWE) activities [4].These works are some of the ongoing efforts to developunderlying services for constructing global sensor net-works.

1http://www.ict-sensei.org2http://www.iot-a.eu/3http://research.microsoft.com/en-us/projects/senseweb/4http://www.opengeospatial.org

0000-0000/0-1900/$00.00 c© 0 – IOS Press and the authors. All rights reserved

2 P. Barnaghi et al. / Sense2Web: A Linked Data Platform for Semantic Sensor Networks

There are currently standards and common frame-works to describe sensor data such as those providedin SWE’s specifications and W3C’s Sensor Ontology5.However, data interoperability in application domainsand data relations to spatial (e.g. geographical location,relative relation in relation to other resources), tempo-ral (e.g. availability, time zone) and thematic (e.g. unitof measurement, type of data, domain) attributes arenot often considered in these specifications. Trust, pri-vacy, security and reliability issues and also resourcediscovery are also issues that need to be considered indesigning access and query mechanisms in large scalesensor networks. To automate data and service provi-sioning in sensor networks there is a need for a com-plete set of solutions, from requirement specificationto discovery, reasoning and integration mechanisms.This paper describes the use of interoperable seman-tic data for this purpose. We discuss publishing datadescriptions for the Semantic Sensor Networks in theform of linked data. This also includes the descrip-tion of a gateway component that acts as an interme-diary for capturing, delivering and presenting dynamicreal world data to consumer applications and users.Collaboration, scalability and semantic interoperabil-ity are the key features in designing large-scale sen-sor networks to support efficient resource distributionand data communication. This generates networked re-sources which collect data from the physical worldas well as data and services on the Web. Interlinkingdata from the physical world and the Web supportsthe provision of networked knowledge [5]. Annotation,processing and reasoning sensor data on a large-scalewill be a challenging task for applications that publishand/or utilise this data from various sources. However,to some extent this is similar to challenges that the se-mantic Web community faces in dealing with the hugevolume of ontologies and semantically annotated datafrom different sources and applications.

Linked data is one way to publish, share and con-nect data via URIs on the Web6. It focuses on inter-connecting data and resources on the Web by defin-ing relations between ontologies, schemas and/or di-rectly linking the published data to other existing re-sources on the Web. The process can be done manuallyor (semi-) automatic mechanisms can be used to createthe links. Publishing data as linked-data enables other

5http://www.w3.org/2005/Incubator/ssn/wiki/Incubator_Report6http://linkeddata.org/

related data and relevant information to be discoveredand facilitates interconnection and integration of datafrom different communities and sources. In this paperwe describe our Sense2Web platform.

Sense2Web facilitates the publication of linked sen-sor data and makes this data available to other Webapplications via SPARQL endpoints7. Our main fo-cus in this paper is sensor description data. The sen-sor observation and measurement data is also repre-sented according to the model that we have devel-oped in [6]. However, storing and querying observationand measurement data has other requirements, suchas time-dependency, scalability, freshness and latencythat are beyond the scope of this paper. We have alsoimplemented a mash-up application using data fromSense2Web to demonstrate reasoning and interpreta-tion of linked sensor data and sensor network resourcedescriptions.

The remainder of this paper is organised as follows:Section 2 describes semantic sensor networks. Sec-tion 3 discusses linked data principles and describesLinked Open Data concepts. Section 4 explains linkedsensor data and linking sensor descriptions to exist-ing resources on the Web. Section 5 demonstratesthe Sense2Web platform for publishing and accessinglinked sensor data and demonstrates a mash-up appli-cation using constructed linked data. Section 6 con-cludes the paper and discusses future work.

2. Semantic Sensor Networks (SSN)

Providing interoperable and machine-interpretabledescriptions of sensor, observation and measurementdata and interconnecting the data to other existing re-sources on the Web are fundamental steps in the con-struction of semantic sensor networks. Sheth et al. [7]propose annotating sensor data with spatial, tempo-ral, and thematic semantic metadata and refer to thisas the Semantic Sensor Web (SSW). This approachuses the current OGC and SWE specifications, extendsthem with semantic Web technologies and provides en-hanced machine-interpretable semantic descriptions.The enhanced descriptions facilitate more autonomousaccess to sensor data and the spatial and temporal at-tributes enable enhanced discovery and utilisation of

7http://semanticweb.org/wiki/SPARQL_endpoint

P. Barnaghi et al. / Sense2Web: A Linked Data Platform for Semantic Sensor Networks 3

the represented data. The W3C Semantic Sensor Net-works Incubator Group (SSN-XG)8 has also developedan ontology for describing sensors. SSN-XG providesa state-of-the-art report on current activities in thisarea9 and the proposed Sensor Ontology (i.e. SSN On-tology) defines higher level concepts to describe sen-sor devices, capabilities and associated platforms10.

Associating sensor and sensor network data withother concepts (on the Web) and reasoning the datamakes this information widely available for differentapplications, front-end services and data consumers.Semantics allow machines to interpret links and rela-tions between the different attributes of a sensor de-scription and also other data existing on the Web orprovided by other applications and resources. Utilis-ing and reasoning this information enables the inte-gration of the data on a wider scale, known as net-worked knowledge [8]. This machine-interpretable in-formation (i.e. semantics) is a key enabler for the se-mantic sensor networks.

Fig. 1. Publishing data from sensors and services on the Web

We have developed a framework for publishing sen-sor data descriptions and for linking this data to otherresources on the Web. The semantically enriched datais made accessible at HTTP level and thus is avail-able for business processes and data integration andcollaboration services. Another important aspect isthe interoperability of the data and scalability of theframework. Utilising metadata and semantic annota-tions to describe sensor data and physical world re-sources in general enables different communities to

8http://www.w3.org/2005/Incubator/ssn/9http://www.w3.org/2005/Incubator/ssn/wiki/State_of_the_art_survey10http://www.w3.org/2005/Incubator/ssn/wiki/Report

_Work_on_the_SSN_ontology

exploit emerging data and exchange information andknowledge in a collaborative environment. User an-notated resources and services similar to those em-ployed by social Web and semantic Web, as well ascommon machine-interpretable descriptions and queryinterfaces, are key aspects in designing the framework.

Figure 1 shows an example of the integration of datafrom different sources. The example illustrates a parcelthat is tagged with an RFID tag which is scanned ev-ery time it is loaded or unloaded. The post delivery vanhas a GPS sensor which reports its location and a twit-ter service is deployed to report the status of the parcelto interested twitter followers. In a semantic integra-tion scenario each of these sensors and services needto be able to describe and/or discover what type of in-formation is published and who can use this informa-tion. This includes the sensor or service descriptionsand also the data reported from sensors and services.In this paper we provide gateway components and ser-vices to make the sensor data accessible to higher levelapplications and/or services and describe how the datais published as linked data in the Sense2Web platform.We also discuss how a consumer service/applicationcan query and access this data using standard querymechanisms. The following describes a gateway com-ponent that enables Sense2Web to access sensor dataand connects capillary sensor networks to higher levelIP-based networks. The gateway component representssensor data and the sensor descriptions via Web ser-vices that can be discovered and utilised in higher levelservices and applications.

2.1. Gateway design for Semantic Sensor Networks

We describe a gateway component for semantic sen-sor networks which connects the capillary sensor net-works to higher application layers. In our design thesensor nodes are connected to the nearest gatewaywhich runs on a more powerful machine. The gatewaymanages the connected sensor nodes provides accessto sensor data through a service layer. The gatewayruns on capable devices (often with persistent powersupply) and supports several connection types such asthe IEEE 802.15.4 compatible communications. Thegateway component provides mechanisms for sensornodes to handshake with available gateway compo-nents and to register their specification and capabili-ties in the form of node context information. The nodecontext includes information such as time zone, avail-


ability, location, measurement capabilities and range.

Fig. 2. Gateway component for Semantic Sensor Networks

The gateway component is divided into three mainlayers. The layers and main interactions are shown inFigure 2. The connectivity layer establishes the con-nection with a capillary sensor network. A connectormodules for each supported sensor/protocol platformshould be developed. External nodes directly connectto the gateway or use multi-hop connections. Thegateway provides a common interface which higherlevel applications and services can access the under-lying sensor networks and their capabilities. When anew node is activated, the node context informationis stored in the gateway repository. The goal of stor-ing this information is to obtain a semantic descriptionthat other processes can exploit and infer the status andcapabilities of each node. The information processinglayer uses the data from the connectivity layer andfrom the context information to support query analy-sis and processing. To establish a reliable connectionbetween sensor nodes and gateway, a method similarto the association and negotiation protocol in the IEEE802.11 Standard’s negotiation steps 11 is developed.The details of the gateway architecture and the negoti-ation mechanism is discussed in [9].

In Figure 3 a mapping of the different layers in thegateway to the OSI model is illustrated. The connec-tivity layer in the gateway provides device connectiv-ity via physical, datalink and network layers to a uni-fied access layer. In the current work we support con-nectivity for the IEEE 802.15.4 capable nodes (the Op-erating system (e.g. SunSpot nodes) or in the protocolstack (i.e. TinyOs 6loWPAN) level support). Incom-ing Queries are processed in the Information Process-ing Layer. The Application Layer in the OSI model is

11http://standards.ieee.org/about/get/802/802.11.html

mapped to the Service Layer in the gateway design.The gateway provides Web service interfaces to rep-resent the sensor nodes and their capabilities and dataand to integrate the network data into higher level ap-plications and services.

Fig. 3. Mapping of the gateway layers to the OSI model layers

2.2. Sensor description model

As an upper level ontology, we use the SSN on-tology from the W3C Semantic Sensor Networks In-cubator Group and extend it to represent context in-formation for the sensor devices in the gateway. Theextended ontology represents information related tosensor instances and their deployment attributes. Eachtime a new sensor is associated to the gateway, a tem-plate for the sensor type is loaded and filled with the in-formation that are the relevant to the node. We assumethat the gateway plug-in developer for a particular sen-sor type will provide the template. If the template didnot exist, the sensor node should send semantically an-notated data according to the designated model as apart of handshaking and negotiation process. The SSNontology allows to define attributes related to mea-surement capability of a sensor node, device features,and platform related attributes. The model extensionincludes the sensors context information. We intro-duce a new Entity "‘SensorContext"’ in the SSN on-tology. SensorContext describes the sensors context interms of available battery, deployment environment at-tributes, location, time and the attributes that can bespecified for the sensor measurement capability in theSSN ontology. The following illustrates a snippet of asensor context description (or in fact this is an instan-tiation of the the SSN schema ontology) 12.

In the following section we describe the concepts oflinked data and discuss how different sensor data de-scriptions are supported and published in the platform.

12http://purl.oclc.org/net/unis/ssnExtended.owl


Fig. 4. A snippet of the device context description

3. Linked data

Publishing data on the semantic Web with machineinterpretable representations facilitates more struc-tured and efficient access to the resources; however se-mantic descriptions without being linked to other ex-isting data on the Web would mostly be processed lo-cally and according to the domain descriptions (i.e.domain ontologies). Linking data to other resourceson the Web enables obtaining more information acrossdifferent domains. The linked data concept was ini-tially introduced by Tim Berners-Lee in 2006 [10].Berners-Lee suggested four main principles to publishlinked data:

– using URIs as names for data,– providing HTTP access to those URIs,– providing useful information for URIs using the

standards such as RDF and SPARQL,– Including links to other URIs.

Publishing annotated and interconnected data is theunderlying principal of creating linked Web resourcesthat is referred to as the Web of Data [10]. The Webof Data can be browsed and accessed as traditionalHTML pages on the Web. However, in the Web ofData instead of HTML links between the pages, theresources are connected via links that can be queriedand interpreted using discovery and search agents [11].Linked data enables users to navigate between dif-ferent data sources by following the data connectionlinks. This allows the linked data consumers to startwith one data source and then browse through a vastnumber of resources interconnected by machine inter-pretable links (e.g. RDF links).

The Web of Data is supported by the SemanticWeb and in particular the Linking Open Data Commu-nity project in the W3C Semantic Web Education and

Outreach Working Group13. The Linking Open DataCommunity project started in 2007 and as reported inSeptember 2010the data sets that have been publishedand interlinked consisted of 203 sets with over 25 bil-lion RDF triples which are interlinked by around 395million RDF links14. The project includes various opendata sets available on the Web such as Wikipedia15,Wikibooks16, Geonames17, and WordNet18. Figure 5shows the data sets that have been published and in-terlinked by the project19. In practice, the linked datapublished on the Web is RDF data that is accessiblevia HTTP interfaces. Recently some public and gov-ernment organisations data has also been publishedas linked-data; for example the UK government pro-vides linked data in different sectors [12] which can bequeried via SPARQL endpoints20.

Fig. 5. Data sets in the Linking Open Data Community project

Construction of sensor data as linked data enablessensor network providers and data consumers to con-nect sensor descriptions to various data existing on theWeb. Relating sensor data attributes such as location,type, observation and measurement to other resourceson the Web enables integration of physical world dataand the logical world data to create business intelli-gence, create smart environments and support auto-

13http://esw.w3.org/SweoIG/TaskForces/CommunityProjects/LinkingOpenData

14http://esw.w3.org/TaskForces/CommunityProjects/LinkingOpenData/DataSets/Statistics

15http://www.wikipedia.org/16http://www.wikipedia.org/17http://www.geonames.org/18http://wordnet.princeton.edu/online/19source: http://richard.cyganiak.de/2007/10/lod/lod-datasets

_2010-09-22_colored.png20http://data.gov.uk/sparql


mated decision making systems among many other ap-plications. The linked sensor data can be queried, ac-cessed and reasoned based on the same principles thatapply to linked data. This creates an open platform topublish and consume interoperable sensor data.

4. Linked sensor data

Sheth et al. in [7] define semantics of sensor Webwithin Space, Time, and Theme attributes. There havebeen different approaches to provide semantic modelsfor each of these attributes independently or in rela-tion to sensors. Some of the common ontologies areSIMILE location ontology21, DAML location ontol-ogy22 for spatial attributes, OWL time ontology23 fortime and common ontologies and vocabularies suchas CyC24, DBpedia25 for thematic data. In a previouswork, we described annotating sensor observation andmeasurement data according to these attributes [13].In this paper, we discuss publishing linked sensor dataand association of this data to the existing resourcesand especially those that are currently a part of theLinked Open Data.

4.1. Spatial attributes

Location specific information for sensors could in-clude specific geographical location attributes definedas longitude and latitude and/or high level conceptsthat describe the location in different terms and in re-lation to other domain concepts (e.g. postcodes). In theSWE Sensor Observation Service (SOS) [4] standardto provide sensor observation and measurement data,the descriptions include location attributes that are ex-plained using GML26 elements. Patni et al. [14] de-scribe a linked sensor data platform that uses locationattributes in the OGC standards and associate them tohigh level concepts using GeoNames and LinkedGeo-Data27 resources. The location concept could be spec-ified in different levels of granularity. It could be a de-tailed specification of a room or a corridor in a build-ing in a detailed level and on a higher level it could re-fer to a campus, a city and so on. The main challenge

21http://simile.mit.edu/2005/05/ontologies/location22http://www.daml.org/experiment/ontology/location-ont23http://www.w3.org/TR/owl-time/24http://cyc.com/cyc/opencyc25http://dbpedia.org/26http://www.opengeospatial.org/standards/gml27http://linkedgeodata.org

for describing sensor location data is how to provide ahigh level of granularity on a global scale without re-quiring to create the world-scale models. For applica-tions with limited scope it is possible to define a lo-cation model and populate it with domain instances.However on a global scale the location instances couldrefer to potentially endless location data. To addressthis challenge, Sense2Web uses two location attributesto describe location of the sensors. The first attributerefers to an instance of a local ontology which is amodel of the current location that a sensor or a nodeis deployed in. This could include higher granularityand detailed information of the physical location suchas rooms, corridors, floors and buildings in an environ-ment. This ontology could be populated and used indifferent applications. We have defined a basic schemafor such an ontology which is discussed in Section 5;however, the sensor data publisher may opt to use adifferent ontology, and as long as the schema link def-initions are available to the data consumers this willnot affect query and accessing of the linked sensor datain the platform. The second location attribute is se-lected from high level location concepts that are avail-able via the Linked Open Data concepts. In particu-lar, we have used concepts referring to different loca-tions selected from DBpedia and GeoNames. Figure 6shows a SPARQL snippet to query a sample locationfrom DBpedia.

Fig. 6. A sample location query from DBpedia

4.2. Temporal attributes

Temporal attributes in sensor data are those de-scribing attributes such as time zone and measure-ment time-stamp. In this context, using common on-tologies for temporal specifications enable linked dataconsumers to query and access temporal features ofdata using standard models and interfaces. There arealso aspects such as freshness for querying the obser-vation and measurement data or availability time forthe sensor and network nodes that can be specified us-ing temporal attributes. This attributes can be then pro-cessed and inferred by application and service level re-


sources to access the appropriate data or to can be usedfor service integration and planning purposes (i.e. byknowing when services will be available or to use it inservice/data compensation processes).

4.3. Thematic attributes

Thematic data provides links between sensor dataand domain knowledge in different applications. Theattributes such as sensor type, description tags, typeof observation measurement, features of interest andother more specific attributes including operationaland deployment attributes describe sensors with do-main knowledge. We use an ontology to provide morespecific sensor descriptions such as those proposed inthe SENSEI project [15] and device specific informa-tion according to the W3C’c Semantic Sensor Net-works Ontology as described in Section 2.2. The highlevel concepts to specify sensor data are provided byreferring to the Linked Open Data resources availableon the Web. For this purpose, we use DBpedia re-sources to define sensor types (i.e. general types) andtags (general concepts and applications) and high levellocation descriptions are adopted from DBPedia andGeoNames. It is also worth mentioning that in manyapplications relying only on general sensor type defini-tions provided by community-driven vocabularies suchas DBpedia will not be sufficient; however, in this pa-per we only demonstrate how linked sensor data canbenefit from existing resources. Similar to location in-formation, for more specific information, local and ap-plication/domain specific ontologies can be used to de-scribe detailed attributes more precisely.

4.4. Sensor specific attributes

Sensor data does not only consist of spatial, tem-poral and thematic features. The sensor as a deviceand the sensing process have also more specific at-tributes and features. In addition to providing links be-tween sensor attributes and other resources, there areapproaches to annotate and link sensor observation ser-vices and also describe device dependent and processspecific features of sensor data. In this context, Janow-icz et al. [16] describe a semantic enablement layeron top of the SWE standards. Henson et al. [17] dis-cuss construction of a Semantic Sensor ObservationService based on the SWE standards. There are severalapproaches that provide ontology based description formore specific sensor data such as those described in[2], [3], and [18]. There are also important issues such

as search, discovery and fusion of sensor data and dy-namic updates that are not in the scope of this paper.The SENSEI project white-paper provides a review onsome of these issues and discusses possible solutionsfor some of these issues[19].

4.5. Re-visiting linked-data principles to publishlinked sensor data

The four main principles described by Berners-Lee[10] are not mandatory guidelines to publish linkeddata; however, following these principles makes thelinked data easily and efficiently accessible to con-sumers on the Web. In this section we revisit the guide-lines and discuss them for publishing linked sensordata.

4.5.1. URIs and Naming:By assigning URIs to sensor network descriptions,

each sensor and/or node has a unique URI that refers toits descriptions. The naming of sensors can follow sim-ilar conventions that are used in HTML pages (or otherresources on the current Web). The SENSEI projectproposes a more specific guideline to define a Univer-sal Resource Name (URN) for sensors. In the SEN-SEI project, the unique resource identifier includes ad-ministrative domains as the first part after the names-pace identifier and then adds resource identifiers to theURN [19]. The following shows a sample URN usingthe SENSEI naming convention.

urn:sensei:surrey.ac.uk:TeloSBSensorTS1:Temperature:SampleRate:3223a-86bca-0123-e123

All resources in SENSEI are identified by a domainand the unique resource identifier is constructed usingthe domain name, sensor type and an internal uniqueidentifier. We propose a similar approach for definingsensor identifiers; with a difference that in our linkedsensor data the identifiers are defined as URIs. This toa large extent addresses the naming issues discussed indeployment and utilisation of large sensor networks. Inthis scenario all the resources are identified by a uniqueURI and access to the low level sensor nodes in thesensor networks are provided via the gateway compo-nents. The sensor network data (i.e. node and devicestatus updates) and observation and measurement datacan be then provided via HTTP level Web services andthe sensor node description and attributes are availableby referring to the linked description data (i.e. the nodedescriptions and status updates are not in the scope ofthis work and are not discussed in this paper).


4.5.2. Providing HTTP access:The linked sensor data can be made available

through HTTP access by simply publishing descrip-tions as Web documents or in a more efficient way, aslinked data approach suggests, by providing SPARQLendpoints to query and access the data. Sensor ob-servation and measurement data can also be madeavailable through HTTP interfaces via sensor obser-vation services. The SWE Sensor Observation Ser-vice (SOS) [4] defines a standard Web service inter-face for requesting, filtering, and retrieving observa-tions and sensor system information. In [20] a serviceoriented middleware architecture for providing HTTPaccess to sensor observation and measurement data isalso discussed. In Section 2.1, we discussed our gate-way component that provides HTTP level access tothe sensor data. The sensor node descriptions are alsopublished as RDF data and we provide SPARQL end-points and standard query interfaces to access this data.Sense2Web also allows users to publish RDF sensordescriptions according to other ontologies and mod-els. The details of publishing RDF sensor data are de-scribed in Section 5.

4.5.3. Providing meaningful descriptions:The sensor data in Sense2Web includes RDF de-

scriptions according to ontologies designed to repre-sent different features and attributes. We propose a twolayer sensor data annotation. This enables sensor dataannotation according to the above described modelsand also using some of the features to link the descrip-tions to other resources on the Web (and in particularLinked Open Data concepts). In Sense2Web, an RDFdescription captures basic attributes of a sensor and/ora node (i.e. spatial and thematic data) and uses publiclyavailable linked data to create links to other resources.

4.5.4. Linking to other URI’s:To describe sensor data the instances for each at-

tributes (i.e. property values) wherever it is applica-ble can be chosen from publicly available linked data.This enables construction of sensor descriptions thatare already linked to other resources based on differ-ent features. We also propose using local ontologiesand vocabularies to provide more specific descriptionsand also allow users to add and associate existing RDFdata, according to other ontologies, to the sensor de-scriptions. Including all this data and publishing it aslinked RDF data creates a set of resources that some oftheir attributes are already described using other Webresources. This allows browsing and accessing the databy referring to different attributes. It also establishes

a link between other RDF descriptions of the sensordata and the high level concepts in the semantic sensornetwork framework.

5. Sense2Web platform

Sense2Web enables publishing linked sensor dataaccording to the four main principles discussed in theprevious section. It enables users to publish their sen-sor description data as RDF triples, use other exist-ing RDF sensor description data, link the descriptionsto the existing resources on publicly available linkeddata repositories and make it available to the data con-sumers through SPARQL endpoints.

Fig. 7. Sensor data publication interface

Figure 7 shows the user interface for publishing anew sensor descriptions in Sense2Web. We use JenaAPI [21] to query DBpedia and other resources to ob-tain values for location, type and descriptive proper-ties. We query the Linked Open Data resources and se-rialise the results using AJAX technology [22]directlyto the page; so user can type a keyword and obtain rel-evant suggestions from on-line repositories. Figure 8shows suggestions from DBpedia for a sample query,"‘Guildford"’ as a location attribute.

The submitted variables are stored in XML formatand Extensible Stylesheet Language Transformations(XSLT)28 is used to transform the submitted data toRDF form. This makes generation of the RDF dataflexible and less dependent to the current model. Weconstruct the RDF data for sensor node descriptionsaccording to a basic RDF structure that captures mainproperties of available sensors and link it to other RDFfiles that provide more specific properties according tocommon sensor ontologies. However, by using a dif-

28http://www.w3.org/TR/xslt20/


ferent stylesheet data can be transformed to anotherformat or other namespaces based on different appli-cations and requirements.

It should be noted that the properties identifiedin main the Sense2Web RDF descriptions are neithercomplete nor fixed. The system includes primary at-tributes to demonstrate the feasibility of publishinglinked sensor data and it can be extended to includemore specific attributes to describe sensor networkdata.

Fig. 8. Resource suggestions from DBpedia for a sample keyword

Fig. 9. Main components in Sense2Web

We use SDB29 a SPARQL database for Jena to storethe RDF triples. To provide SPARQL endpoints, weuse an open source SPARQL server for Jena calledJoseki30. We have also implemented interfaces that en-able users to obtain the query results in a different for-mat such as XML, RDF and SPARQL protocol for-mat31.

29http://openjena.org/SDB/30http://joseki.sourceforge.net/31http://www.w3.org/TR/rdf-sparql-protocol/

Figure 9 shows the main components for linked datapublication in the Sense2Web platform and the inter-faces to access the published data.

Fig. 10. Related data for a sample resource (i.e. "‘Guildford"’) fromDBpedia

5.1. Mash-up application

Fig. 11. A sample mash-up application

To demonstrate the linked data usage and integrationof data from different sources we have created a mash-up application using Google Maps API32. For this ap-plication, we use the location attributes and retrieve ge-ographical coordinates of the nodes by processing theLinked Open Data resource descriptions. The applica-tion retrieves related properties of the resources fromthe linked sensor data repository and lists availablesensors and their properties through a Google Maps

32http://code.google.com/apis/maps/


application. In the current demo, we only retrieve pub-lished nodes and show them in a map overlay. Thiscan be extended to discovering other related resourcessuch as nearby locations, districts and other related in-formation available through browsing different links.Figure 10 shows some of the related data available forour sample query ("‘Guildford"’) from DBpedia.

Figure 11 demonstrates the application and showsavailable information for a sample sensor publishedaccording to our description model. In our sample, wecreate links between domain (i.e. thematic) level at-tributes of sensor data, type and keyword descriptions.In fact, if more public semantic data related to differentattributes of sensor description is available, then moreproperties of sensor description can be linked to otherresources. This will require more common ontologiesto describe sensor types, sensor platforms, sensor mea-surement attributes, sensor devices, etc.

6. Conclusions and future work

This paper discusses construction of linked datafrom sensor data and in particular providing sensordescriptions and associating their attributes with re-sources on the Web. The presented work providesinteroperable sensor data descriptions and facilitatespublication of this data as linked data. We describe ourplatform and discuss how existing Linked Open Dataresources are used to describe spatial and thematic at-tributes of the sensor descriptions and to link them tothe Web data. We discuss using stylesheets to trans-form the annotation data to different designated sensordescription formats such as W3C SSN ontology rep-resentations. The paper describes the linked data prin-ciples and we describe the implemented prototype. Ademonstration of the framework following the linkeddata principles and using a mash-up application is alsoprovided.

The future work will focus on defining and inte-grating different sensor platforms to the system. Thiswill include providing sensor and service discoverymechanisms and evaluating the scalability of the plat-form. Providing reasoning methods and processing thelinked data descriptions to integrate sensor data intobusiness applications and smart environments will alsobe a part of the future work. We plan to define descrip-tion models for network nodes and servers to facilitateautonomous management of resources in the sensornetworks and to represent these descriptions as linkeddata in the platform.

Acknowledgements

This paper is an extension of the "Publishing LinkedSensor Data" paper published in proceedings of the3rd International Workshop on Semantic Sensor Net-works (SSN) in November 2010. The work describedin this paper is partially undertaken in the context ofthe IoT-A project, IoT-A: Internet of Things - Archi-tecture (http://www.iot-a.eu/public) contract number:257521.

References

[1] J.-H. Kim, H. Kwon, D.-H. Kim, H.-Y. Kwak, and S.-J. Lee,“Building a service-oriented ontology for wireless sensor net-works,” Computer and Information Science, 2008. ICIS 08.Seventh IEEE/ACIS International Conference on, pp. 649–654,May 2008.

[2] D. Russomanno, C. Kothari, and O. Thomas, “Sensor ontolo-gies: from shallow to deep models,” System Theory, 2005. SSST’05. Proceedings of the Thirty-Seventh Southeastern Sympo-sium on, pp. 107–112, March 2005.

[3] M. Eid, R. Liscano, and A. El Saddik, “A universal ontol-ogy for sensor networks data,” Computational Intelligence forMeasurement Systems and Applications, 2007. CIMSA 2007.IEEE International Conference on, pp. 59–62, June 2007.

[4] SWE, 2008, botts, M. and Percivall, G., and Reed, C. andDavidson, J., OGC Sensor Web Enablement: Overview andHigh Level Architecture, The Open Geospatial Consortium,http://portal.opengeospatial.org/files/?artifact_id=25562.

[5] M. Hauswirth and S. Decker, “Semantic reality - connectingthe real and the virtual world,” Microsoft SemGrail Workshop,pp. 21–22, June 2007.

[6] P. M. Barnaghi, S. Meissner, M. Presser, and K. Moessner,“Sense and sens’ability: Semantic data modelling for sensornetworks,” in Proceedings of the ICT Mobile Summit 2009,June 2009.

[7] A. Sheth, C. Henson, and S. Sahoo, “Semantic sensor web,”Internet Computing, IEEE, vol. 12, no. 4, pp. 78–83, July-Aug.2008.

[8] S. Decker and M. Hauswirth, “Enabling networked knowl-edge,” in CIA ’08: Proceedings of the 12th international work-shop on Cooperative Information Agents XII. Berlin, Heidel-berg: Springer-Verlag, 2008, pp. 1–15.

[9] F. Ganz, P. Barnaghi, F. Carrez, and K. Moessner, “Context-aware management for sensor networks,” in Proceedings ofthe Fifth International Conference on COMmunication SystemsoftWAre and middlewaRE (COMSWARE11), June 2011.

[10] T. Berners-Lee, “Linked data,” 2006, http://www.w3.org/DesignIssues/LinkedData.html.

[11] C. Bizer, T. Heath, K. Idehen, and T. Berners-Lee, “Linked dataon the web (ldow2008),” in WWW ’08: Proceeding of the 17thinternational conference on World Wide Web. New York, NY,USA: ACM, 2008, pp. 1265–1266.

[12] J. Sheridan and J. Tennison, “Linking uk government data,” inWWW2010 workshop:Linked Data on the Web (LDOW2010).ACM, 2010.


[13] W. Wei and P. Barnaghi, “Semantic annotation and reasoningfor sensor data,” in EuroSSC’09: Proceedings of the 4th Euro-pean conference on Smart sensing and context. Berlin, Hei-delberg: Springer-Verlag, 2009, pp. 66–76.

[14] J. Pschorr, C. Henson, H. Patni, and A. Sheth, “Sensor discov-ery on linked data,” in Proceedings of 7th Extended SemanticWeb Conference, 2010.

[15] C. Villalonga, M. Bauer, V. Huang, J. Bernat, and P. Barnaghi,“Modeling of sensor data and context for the real world in-ternet,” in 7th IEEE Workshop on Context Modeling and Rea-soning (CoMoRea) at the 8th IEEE International Conferenceon Pervasive Computing and Communication (PerCom’10).IEEE, 2010.

[16] A Transparent Semantic Enablement Layer for the GeospatialWeb, Washington DC, USA, 2009.

[17] C. A. Henson, J. K. Pschorr, A. P. Sheth, and K. Thirunarayan,“Semsos: Semantic sensor observation service,” CollaborativeTechnologies and Systems, International Symposium on, vol. 0,

pp. 44–53, 2009. [Online]. Available: http://dx.doi.org/10.1109/CTS.2009.5067461

[18] H. Neuhaus and M. Compton, “The semantic sensor networkontology: A generic language to describe sensor assets,” in AG-ILE 2009 Pre-Conference Workshop Challenges in GeospatialData Harmonisation, 2009.

[19] “Sensei project, the sensei real world internet architecture,white-paper,” 2010, http://www.sensei-project.eu/.

[20] H. Abangar, P. Barnaghi, K. Moessner, R. Tafazolli, A. Nnae-mego, and K. Balaskandan, “A service oriented middlewarearchitecture for wireless sensor networks,” in Future Networkand Mobile Summit 2010, 2010.

[21] Jena, 2008, jena, A Semantic Web Framework for Java, http://jena.sourceforge.net/.

[22] J. J. Garrett, “Ajax: A new approach to web applications,”2005, http://www.adaptivepath.com/ideas/essays/archives/000385.php.

ios press sense2web: a linked data platform for semantic … · undeﬁned 0 (0) 1 1 ios press...

Documents