Using a Hybrid Semantic Similarity Assessment Model to Resolve Semantic Heterogeneities in SDIs Case Study: Iranian Water and Wastewater Company

Many countries aim to design and build Spatial Data Infrastructure (SDI) to facilitate, manage and share spatial data. Different public or private organizations provide data sources in diverse ways and various contextual situations such as weather conditions, coordinate system definitions or acquisition times. Therefore, SDI should be semantic-based, as possible as it can, to deal with different user languages, requirements. Such an SDI can help providing appropriate representation and search. Since the data integration is an essential part of each information system, semantic similarity is getting more attention in the web world. An efficient spatial data sharing across different organizations is considered to have significant contributions to the sustainable development of today’s communities. As the quantity and accessibility of spatial data is tremendously increasing via web, interpreting, handling and retrieving of this data has become a difficult task. The data suppliers come from various information communities with differing conceptualizations of the world. So, this data is heterogeneous in essence and distributed over several sources. Since the acquisition of geospatial data is extremely expensive, developing mechanisms for reusing and sharing geographic information are necessary to save costs. Besides, customer orientation and personalization of data sources is central to enable flexible and multipurpose usage of the data and to provide customers with the required data. Ordinary information retrieval systems are limited to syntactic retrieval mechanisms and therefore cannot deal with semantic differences in the customer's and the data supplier's conceptualization. The Open Geospatial Consortium (OGC) has established standards for storing, discovering, and processing geographical information but these standards cannot solve the semantic problem. Today, the semantic heterogeneity is considered as the main obstacle to the full interoperability among spatial data sources. Geospatial data describes real world geographic features by their spatial extent and their location. Hence, properties are necessary to capture the semantics underlying geospatial data, because they can represent spatial qualities such as shape. The notion of semantic similarity serves as an indicator for relevance in the retrieval process.
This paper uses an ontology-based approach and description logic to resolve the semantic heterogeneity. For this purpose, semantic similarity measurement is used to interpret, handle, and retrieve data in terms of semantically similar concepts. In order to calculate similarities, two existing similarity measurement models were combined: Feature model and Network model. While Feature model computes similarity of concepts based on their common and distinctive properties, Network model puts the concepts in a semantic network and computes the similarity based on the relations of the concepts in the network. This paper proposes a hybrid similarity model as a computational model for semantic similarity measurement. This hybrid model enables the necessary expressiveness to capture semantics underlying geospatial data. The shortcomings and benefits of each model with respect to the requirements of semantic information retrieval of geospatial data are described. Retrieval systems use similarity measures to determine the relevance. Only a retrieval system which returns cognitively adequate results can successfully support human users. The proposed model retrieves relevant information by measuring the semantic similarity of concepts to a given query. The methodology has been tested on some parts of Iranian Water and Wastewater Company’s infrastructure as a case study. Since semantic similarity is an appropriate means to resolve semantic heterogeneity in retrieving data in SDIs, the proposed model can help users by representing similarity in a quantitatively manner. This paper has considered blockage in pipeline as user search concept. The results of similarity represent the advantages of the proposed model. In addition, the results showed that the most similar concept to user search concept was Elbow with %42.5 similarity because of its curvature.