Abstract

Access to geographic information has radically changed in the past decade. Previously, it was a specific task, for which complex desktop Geographic Information Systems (GISs) were built, and geographic data was maintained locally, managed by a restricted number of technicians. With the significant impact of the world-wide-web, an increasing number of different geographic services became available from heterogeneous sources. To support interoperability among different providers and users, GIS agencies have started to adopt Spatial Data Infrastructure (SDI) models. Usually, each SDI service provider publishes and gathers geographic information based on its background knowledge. Hence, discovering, chaining, and using services require a semantic interoperability level between different providers. This problem is typically referred as the need for 'semantic interoperability among autonomous and heterogeneous systems' and it is a challenge for current SDIs, due to their distributed architecture. This thesis provides a framework to approach the semantic heterogeneity problem in the field of geo-services - services that deal with the generation and management of geographical information - among distributed SDIs. The framework is based on: (i) a peer-to-peer (P2P) view of the semantics of web service coordination, implemented by using the OpenKnowledge system and (ii) the use of a specific semantic matching solution called Structure Preserving Semantic Matching (SPSM). SPSM is a basic module of OpenKnowledge as it enables web service discovery and integration by using semantic matching between invocations of web services and web service descriptions. We applied the OpenKnowledge system on a realistic emergency response scenario and selected SDI services. We modeled an emergency response scenario, i.e., a potential flooding event in the area of Trento. The scenario is based on the past experience and actual emergency plans as collected from interviews with personnel of the involved institutions and from related documents. Within this emergency response scenario a detailed implementation of selected SDI services is presented, namely the gazetteer, map and download services. The SPSM solution has been assessed on a set of GIS ESRI ArcWeb services. Two kinds of experiments have been conducted: the first experiment includes matching of original web service signatures with synthetically altered ones. In the second experiment a manual classification of the GIS dataset has been compared to the unsupervised one produced by SPSM. The evaluation results demonstrate robustness and good performance of the SPSM approach on a large (ca. 700.000) number of matching tasks. In the first experiment a high overall matching relevance quality (F-measure) was obtained (over 55%). In the second experiment the best F-measure value exceeded 50% for the given GIS operations set. SPSM performance is good, since the average execution time per matching task was 43 ms. This suggests that SPSM could be employed to find similar web service implementations at runtime. The aforementioned results suggest the practical real time application of the SPSM approach to: (i) discovering geo-services from specific geographic information catalogs, (ii) composing specific geo-processing services, (iii) supporting coordination of geo-sensor networks, and (iv) supporting geo-data discovering and integration.