Title:FunMap: Efficient Execution of Functional Mappings for Knowledge Graph Creation

Abstract:Data has exponentially grown in the last years, and knowledge graphs constitute powerful formalisms to integrate a myriad of existing data sources. Transformation functions -- specified with function-based mapping languages like FunUL and RML+FnO -- can be applied to overcome interoperability issues across heterogeneous data sources. However, the absence of engines to efficiently execute these mapping languages hinders their global adoption. We propose FunMap, an interpreter of function-based mapping languages; it relies on a set of lossless rewriting rules to push down and materialize the execution of functions in initial steps of knowledge graph creation. Although applicable to any function-based mapping language that supports joins between mapping rules, FunMap feasibility is shown on RML+FnO. FunMap reduces data redundancy, e.g., duplicates and unused attributes, and converts RML+FnO mappings into a set of equivalent rules executable on RML-compliant engines. We evaluate FunMap performance over real-world testbeds from the biomedical domain. The results indicate that FunMap reduces the execution time of RML-compliant engines by up to a factor of 18, furnishing, thus, a scalable solution for knowledge graph creation.