A CONCEPTUAL FRAMEWORK FOR QUALITY ASSES SMENT OF SEMANTIC MAPPING BETWEEN SPATIAL ONTOLOGIES Mohamed Bakillah a , Mir Abolfazl Mostafavi a , Yvan Bédard a , Jean Brodeur b a CRG, 0611 Pavillon Casault, Département des Sciences Géomatiques, Université Laval, Québec, Canada, G1K 7P4- Mohamed.bakillah.1@ulaval.ca, (Mir-Abolfazl.Mostafavi, Yvan.Bedard)@scg.ulaval.ca b Centre d’information topographique de Sherbrooke, 2144 King, Canada- brodeur@nrcan.gc.ca KEYWORDS: Data Quality, Ontology Mapping, Semantic Integration, Semantic Similarity Model, Mapping Conflict Predicates ABSTRACT: Quality of integrated data depends on the quality of original data sources, but it can also be affected by the semantic mapping process between ontologies of different sources, thus influencing the quality of querying between multiple data sources. Being aware of semantic mapping quality could help interpreting mapping results in order to obtain better integration of heterogeneous data and would provide higher data quality to users. The question that is still unanswered is how semantic mapping quality can be defined and represented. In this paper, we propose a conceptual framework for characterising semantic mapping quality, which includes a metamodel showing relationship between semantic mappings and their quality aspects, as well as original definitions for characteristics of mapping quality. We define several Mapping Conflict Predicates that can be used to detect incoherence between mappings. We also propose a new semantic model of mapping that includes the different characteristics of mapping quality, which we called semantic model of quality mapping. 1. INTRODUCTION Recent advances in spatial information technologies and the increasing number of spatial data sources available to users emphasize the importance of spatial data integration, which becomes even more important in the context of spatial decision making where a fast and effective processing of data is necessary. The quality of integrated data depends on approaches that are used to resolve heterogeneities between data coming from different sources. For semantic integration of heterogeneous geospatial data, many semantic models of mapping were proposed to establish semantic relationships between ontologies describing these sources (Noy and Musen, 2001; Maedche and Staab, 2002; Doan et al ., 2004; Mostafavi, 2006) or between schemas (Do and Rahm, 2001; Madhavan et al., 2001; Berlin and Motro, 2002). Evaluation approaches used to determine the validity of these semantic models of mapping showed that they achieve a variable performance (Do et al., 2003). In fact, they are adapted to specific situations, i.e. different structures of schema or representations of concepts in geospatial databases, etc. Nevertheless, the result of the semantic mapping process has a significant impact on decision making since it takes part in the query processing between multiple sources (Bouquet et al., 2005). Consequently, it can affect the quality of data that will be provided to users. The user who is unaware of the quality of the semantic mapping process is unable to judge the quality of data which results from the semantic integration process. In this paper, we present a conceptual framework where we define and represent semantic mapping quality. This framework identifies and defines the multiple characteristics of mapping quality such as precision, coherence of mapping, etc. We also propose a semantic model of mapping that explicitly includes these characteristics, which we called semantic model of quality mapping. The content of this paper is structured as follow: section 2 gives the motivation of our research. Section 3 is a review of existing research related to ontology mapping and data quality. Section 4 presents our approach and a metamodel for mapping quality. In section 5 we propose the semantic model of quality mapping between ontologies. Section 6 presents the conceptual framework for mapping quality. Section 7 concludes this paper. 2. MOTIVATION Ontology integration is the process of forming an ontology for a given subject by the re-use of several ontologies describing different subjects (Sofia Pinto and Martins, 2001). It generally involves a semantic mapping process, which consists in identifying a formal expression describing the semantic relationship between concepts of different ontologies (Bouquet et al., 2005). It is known that the quality of data cannot be guaranteed following the integration, since it depends on each source (Wand and Wang, 1996). We also argue that the quality of data is affected during the integration process since mappings are used to rewrite a query on a first source for another source (Bouquet et al., 2005). Ontology A Ontology B Ontology Mapping Internal Quality Internal Quality Query Answering External Quality User Model of Mapping quality affects affects affects fitness for use Figure 1: Impact of semantic model of mapping quality on external quality Consequently, the external data quality (i.e. quality perceived by users, also called fitness for use) is affected not only by internal data quality (accuracy, consistency, actuality, etc.) but also by semantic mapping quality (Figure 1). Mapping quality can also play a significant role in interpreting results of the