Quality is an important aspect of data discovery in the Semantic Sensor Web. This work extends current endeavors to make the Sensor Web more semantic by introducing an ontology design pattern which facilitates the modeling of aspects of spatial data quality. The implementation of a software program over two scenarios demonstrates the usefulness of the ontology design pattern for the Semantic Sensor Web.
Two major aspects of quality characterization in the SSW can be distinguished:
quality aspects that have to do with the characteristics of the data sources,
and those that have to do with the creation of applications based on these data
sources. These two categories of quality aspects were already summarily
mentioned in [
Quality is the degree to which a data or service ful lls the needs of a consumer. It is a function of intangible properties (of the data) considered pertinent to the satisfaction of the consumer's needs.
The intangible properties considered pertinent to the satisfaction of the
consumer's needs are also called components of data quality. As regards spatial
data, quality components vary from author to author. For instance, ISO 19113
includes completeness, logical consistency, positional accuracy, temporal
accuracy and attribute accuracy (see [
Gangemi and Presutti [
The pattern presented in this section is obtained by extraction from the Stimulus Sensor Observation (SSO) ontology design pattern 4. Two classes are left out: the stimulus and the sensor. They are not included because for a quality assessment operation, there is no need to describe the stimulus, nor is there a need to describe a sensor that performs a measurement. Instead, it su ces that the data consumer describes the procedure used for quality assessment.
The documentation of the pattern uses the elds suggested in [
Name: ontology design pattern for spatial data quality characterization Intent: to describe spatial data quality components of sensor observations Competency questions: (i) what is the value of the spatial data quality component for this sensor observation? (ii) what is the quality criterion used to assess the quality of the sensor observation? Elements: The pattern has 5 elements: Data, DataQualityCriterion, DataQualityComponent, DataQualityObservation and DataQualityResult. A data is the output of an observation process involving a sensor, a stimulus, a sensed property and a feature. It is equivalent to `Observation' as de ned in the SSN ontology. A DataQualityComponent is any property of the data that the consumer would like to approximate. Examples of spatial data quality components were mentioned in section 2. A DataQualityCriterion is a criterion de ned by the data consumer to get information about the quality of the data. A DataQualityObservation is an operation by which a data quality value is assigned to a data quality component using a data quality criterion.
The outcome of a data quality observation is a DataQualityResult. Diagram: see Figure 1 Consequences: Bene ts: (i) Reasoning and inference of spatial data quality component values for existing sensor observations (ii) Detection of inconsistencies during the integration of observations with di erent quality levels (iii) Detection of inconsistencies during the integration of observations for which di erent quality criterion have been used to assess the spatial data quality. Trade-o s: The pattern does not give a quality value like `high quality' or `low quality' as an end result. Instead, it helps to infer the value of a spatial data component (e.g. resolution = 20m) and it is left up to the data consumer to decide whether `resolution = 20m' means high or low quality for the task at hand.
Known uses: see examples of uses in section 4 Extracted from: the SSO ontology design pattern / the SSN ontology Building block: http://wsmls.googlecode.com/svn/trunk/global/
Patterns/Quality/DataQuality/dataqualitymodule.owl 4
A software program was developed to test the usefulness of the ODP. This
software program serves as a proof that (i) the pattern can be used to perform
inference of quality component values and (ii) the pattern can be used to warn
against the integration of datasets for which di erent quality criteria are used
for quality assessment. This method for validation falls into the category
`Empirical validation' introduced in [
ScenarioA - Inference of data quality value. A decision-maker has at his/her
disposal di erent OGC services delivering data about an observed property. He/she
would like to answer the question - what is the value of the spatial resolution
for an observation o ering? 6 - in order to compare them and select the most
appropriate for his/her task. A quality criterion for the observation o ering is
the sampling density 7 (i.e. number of sensors per square meters).
Comment: Information about the spatial resolution of an observation o ering
can be deduced from the GetCapabilities le of an OGC service. Using the ODP
introduced earlier for this scenario, it is only required to parse the
GetCapabilities le and assert (or store) the spatial extent of the observation o ering as
well as its number of sensors. The spatial resolution can then be inferred using a
SWRL rule (see details in the Java code). Inference in turn is useful to address
one of the drawbacks of adding semantic annotations to sensor nodes in sensor
networks. In e ect, [
ScenarioB - Detection of inconsistencies. This scenario is adapted from the Oil Spill scenario 8 of the European project ENVISION 9. The project aims at developing an infrastructure for environmental web services with ontologies.
Scenario: accidental oil releases to the sea may have severe consequences on both natural resources and human enterprises. For oil spill decision making, it is essential to be able to predict the fate and e ects of the spilled oil. Fate prediction requires data on the spill (location, time, amount, oil type), the environmental conditions (wind, current), and geography (sea depths, coast line). A decisionmaker has di erent datasets for oil spill prediction at his/her disposal and would like to combine them.
Comment: For the purposes of the illustration, it is assumed that - within this scenario - the di erent types of data for oil spill prediction are available in vector format. It is also assumed that the decision-maker has done a preliminary look-over where all the datasets available were found to have a similar spatial resolution (say 100meters).
Given that there are various ways of de ning the spatial resolution for vector data, an additional question to answer is: what criterion is used to assess the spatial resolution of the observation o ering? 10 in order to ensure that heterogeneous datasets which have the same data quality value are e ectively compatible with respect to their resolution. This check is possible with the ODP proposed through a simple SPARQL query (see details in the Java code). 5
Knowledge about the quality of sensor observations is an important aspect of the discovery of resources in the Semantic Sensor Web. This work has suggested an ontology design pattern (ODP) to characterize the quality of sensor observations. The ODP is relevant for the annotation of sensor observations with spatial data quality components. It can be used to infer spatial data quality component values for existing sensor observations and warn against the integration of sensor observations assessed with di erent quality criteria. The ODP was aligned to the foundational ontology Dolce Ultra Light and validated through the development of a software program. 8 See a detailed presentation of the oil spill scenario at http://envision.brgm-rec.
fr/OS-UseCase.aspx (last accessed: August 30, 2012). 9 See a presentation of the project at http://www.envision-project.eu. 10 This question is an application of the competency question (ii) from section 3 to the spatial data quality component `Resolution'. The work is funded by the German Academic Exchange Service (DAAD A/10 /98506), and the European Commission through the ENVISION Project (FP7249170). It is conducted within the International Research Training Group on Semantic Integration of Geospatial Information (DFG GRK 1498). The ideas presented were shaped through discussions with Patrick Maue, Alejandro Llaves, Werner Kuhn, Alkyoni Baglatzi, Henry Michels and Marcell Roth.