Pattern matching for heterogeneous geodata sources using attributed relational graph and probabilistic relaxation

Information retrieval and intelligent search among heterogeneous data sources still continue to be challenging tasks. In this study, an attributed relational graph was employed to model the semantic information of heterogeneous geodata sources. Based on the attributed relational graphs, probabilistic relaxation was employed for pattern matching between different data sources. The initial probability and compatibility coefficients were calculated based on the combined evidence from semi-structured geodata sources and the characteristics of discrete and categorical variables. Experiments on automatic pattern matching were carried out and the results demonstrated the effectiveness of the proposed approach in element mapping between heterogeneous data sources.