Light-weight and adaptive reasoning for mobile web services

2017-01-13T04:17:07Z (GMT) by Steller, Luke Albert
The growth of smart phones and PDAs coupled with the emergence of Web Services as the de facto technology for supporting seamless heterogeneous integration has led to an emerging focus on mobile services. This emergence of mobile services necessitates service selection mechanisms that are accurate and efficient. Service selection involves matching of a user's requirements against the available services in the user's environment. It is well established that accuracy in service matching is improved by the use of semantics as opposed to simpler approaches such as keyword / interface matching. Semantic matching is performed by semantic reasoners. Current approaches to semantics based service selection tend to perform matching using external / remote high performance servers, because reasoning is a computationally complex and resource intensive activity that does not scale well to mobile devices. However, there are several advantages of performing semantic service matching on-board the mobile device. For instance, on-board matching avoids the overheads associated with the provision and maintenance of external servers to perform matching remotely. Additionally, continuous network access to a remote server has been shown to be a relatively higher drain on a mobile device's battery power when compared to processing activities. Furthermore, a connection may not always be available since mobile devices suffer from intermittent connectivity and frequent disconnection. There may also privacy concerns with transmitting sensitive data to a third party remote server (e.g. a user's shopping preferences and habits). Therefore, in this thesis we propose and develop a novel light-weight and adaptive approach for on-board semantic mobile matching. This thesis makes two significant contributions. Firstly, due to computational complexity, current reasoners cannot perform matching of large ontologies on mobile resource constrained devices. Therefore, we propose and develop mTableaux which enables mobile semantic matching by performing optimisations of the well-known Tableaux algorithm that is used in many of the state-of-the-art open source and commercial reasoners today. These optimisations result in improving the computational efficiency of the semantic reasoning process with a specific focus on scaling to mobile devices, without significantly reducing result accuracy. Secondly, current reasoners typically produce only a positive or negative result under an "all or nothing" principle in which the matching task must be completed in full before a result is provided. Therefore, we propose and develop an adaptive and incremental approach to deliver the outputs of a reasoning task. This allows a mobile user to get valid partial results from a reasoner depending on constraints such as changing context, time or availability of computational resources. We have implemented our proposed light-weight and adaptive reasoning strategies, and conducted extensive experimental performance evaluations which clearly demonstrate that our strategies improve response time and enable incremental matching. Our performance evaluations clearly demonstrate that the efficiency improvements in response time do not compromise accuracy. This evaluation includes tests on a resource constrained mobile device and a comparison of our approach against commercial and open source reasoners in desktop environments, using two realistic application scenarios as well as publicly available ontologies. In summary this dissertation has addressed the problem of enabling efficient and accurate mobile reasoning on small devices to meet dynamic resource levels and user needs in mobile environments. The research done over the course of this dissertation has been published in one international journal paper, seven conference papers and one workshop paper.