Content Quality Assessment and Acceptance Testing in Location-Based Services

In this paper, we develop and evaluate an approach to assessing the content quality in a location-based service (LBS). The proposed approach, instead of assessing the quality in absolute terms such as completeness or accuracy, measures the effect that the imperfection of the content is having on the reliability of that specific LBS. We apply the basic ideas from Software Reliability Engineering (SRE), but develop a modification of SRE, 2-Branch, in order to separate content quality from other factors, such as positioning imprecision, and to reduce the measurement error. In our experimental study, we first compare 2-Branch to the standard SRE, after which we experimentally analyze some properties of SRE methodology as such in the context of an LBS. The experiments indicate that 2-Branch has in most cases a lower measurement error than the standard SRE. A corollary to that is that 2-Branch can achieve, therefore, as low an error level as the standard SRE, using a worse and thus cheaper oracle. Getting a good oracle is probably the main cost factor in evaluating the quality of an information service, thus being able to use a cheaper one may result in significant savings.