A Context Quality Model to Support Transparent
Reasoning with Uncertain Context

Susan McKeever, Juan Ye, Lorcan Coyle, and Simon Dobson
System Research Group, School of Computer Science and Informatics
UCD, Dublin, Ireland
susan.mckeever@ucd.ie
Abstract. Much research on context quality in context-aware systems
divides into two strands: (1) the qualitative identification of quality mea-
sures and (2) the use of uncertain reasoning techniques. In this paper,
we combine these two strands, exploring the problem of how to identify
and propagate quality through the different context layers in order to
support the context reasoning process. We present a generalised, struc-
tured context quality model that supports aggregation of quality from
sensor up to situation level. Our model supports reasoning processes
that explicitly aggregate context quality, by enabling the identification
and quantification of appropriate quality parameters. We demonstrate
the efficacy of our model using an experimental sensor data set, gaining
a significant improvement in situation recognition for our voting based
reasoning algorithm.
1 Introduction
The information used by context-aware systems to recognise different contexts
is often imperfect. Sensor data is prone to noise, sensor failure and network
disruptions. Users actions can contribute to degradation of information quality,
such as the failure of users to carry their locator tags. Further uncertainty can
be introduced in the reasoning process, such as the use of fuzzy functions to
quantify vague context or difficulty in defining accurate inference rules [14].
Existing work in the area of context quality focuses on two main areas: (1) The
qualitative identification of context quality parameters, often as part of a context
modelling exercise, such as the work done by [4,5,6]; and (2) the quantitative use
of reasoning techniques that incorporate context uncertainty such as Bayesian
networks [10] and fuzzy logic [7].
The qualitative work provides a useful vocabulary for identifying and mod-
elling context quality. However, such measures are usually associated with ’con-
text’, without specification of quality for each layer of context. Quality issues
for low level sensor data are different from those at higher levels of context and

This work is partially supported by Enterprise Ireland under grant number
CFTD 2005 INF 217a, and by Science Foundation Ireland under grant numbers
07/CE/I1147 and 04/RPI/1544.
K. Rothermel et al. (Eds.): QuaCon 2009, LNCS 5786, pp. 65–75, 2009.
c© Springer-Verlag Berlin Heidelberg 2009

66 S. McKeever et al.
a context quality model must reflect this [9]. Also, the aggregation of quality
across the layers must be addressed in order to produce a meaningful and use-
able indicator of context quality to applications. This aggregation will support
reasoning schemes that can propagate uncertainty from sensor level upwards. For
example, Dempster Shafer [13] or voting algorithms [2] for context reasoning can
incorporate explicit quantification of uncertainty of context sources.
This paper presents a UML-based structured model of context quality for each
layer of context. We also include an aggregation model that contains a general
set of quality measures and their propagation across context layers. Designers of
context-aware systems can use our combined models to (1) identify and model
context quality issues and (2) to support the specification of quality aggregation.
In particular, context-aware systems using transparent reasoning techniques that
aggregate quality from sensors upward will benefit from our modelling approach.
We demonstrate our work by generating quality parameters for an experimental
dataset. We incorporate these quality parameters into a voting-based reasoning
algorithm. Our results show that situation recognition is significantly improved
with the inclusion of our modelled context quality than when quality is not used.
This remainder of this paper is organised as follows: Section 2 describes re-
lated work by other researchers; Section 3 details our structured quality and
aggregation models and their relevance to context reasoning schemes; In Section
4, we demonstrate and critique our model with an experimental dataset. Finally,
in Section 5, we conclude our work and define our future research direction.
2 Related Work
Previous work on modelling context quality provides various well documented
parameters for context quality, such as context confidence [3,10,11] to indicate
probability of correctness and freshness [1,3,4,8] to indicate the degradation of
information over time. Lower level sensor quality measures such as precision,
accuracy and resolution [1,4] are used to define sensor data issues. Such work
provides useful semantics for exploring the nature of context quality issues. Other
modelling approaches include placeholders for quality parameters within struc-
tured models of context. For example, Henricksen and Indulska’s [6] Object Role
Modelling context model associates context facts with zero or more quality pa-
rameters and associated metrics. Similarly, Gu et al. [5] describe a context model
that includes a quality ontology with specific parameters and metrics. They in-
clude a set of commonly used parameters in their ontology. Both of these mod-
els use similar modelling constructs for quality. However, sensor and situation
quality parameters are not separately identified.
The modelling approaches described do not model or aggregate quality
parameters at each layer of context. We address this as follows: (1) We pro-
vide a structured (UML) extendable context quality model that includes qual-
ity parameters for sensor, abstracted context and situations, as illustrated in