A multi-faceted method for science classification schemes (SCSs) mapping in networking scientific resources

Science classification schemes (SCSs) are built to categorize scientific resources (e.g. research publications and research projects) into disciplines for effective research analytics and management. With the explosive growth of the number of scientific resources in distributed research institutions in recent years, effectively mapping different SCSs, especially heterogeneous SCSs that categorize different kinds of scientific resources, is becoming an increasingly challenging problem for facilitating information interoperability and networking scientific resources. To effectively realize the heterogeneous SCSs mapping, we design a novel multi-faceted method to measure the similarity between two classes based on three important facets, namely descriptors, individuals, and semantic neighborhood. Particularly, the proposed approach leverages a hybrid method combining statistical learning, semantic analysis and structure analysis for effective measurement with the exploitation of symmetric Tversky's index, WordNet dictionary and the Hungarian Algorithm. The method has been evaluated based on two main SCSs that need mapping for information management and policy-making in NSFC, and shown satisfying results. The interoperability among heterogeneous SCSs is resolved to enhance the access to heterogeneous scientific resources and the development of appropriate research analytics policies.