FlexRML: A Flexible and Memory Efficient Knowledge Graph Materializer

We present FlexRML, a flexible and memory efficient software resource for interpreting and executing RML mappings. As a knowledge graph materializer, FlexRML can operate on a wide range of systems, from cloud-based environments to edge devices, as well as resourceconstrained IoT devices and real-time microcontrollers. The primary goal of FlexRML is to balance memory efficiency with fast mapping execution. This is achieved by using C++ for the implementation and a result size estimation algorithm that approximates the number of N-Quads generated and, based on the estimate, optimizes bit sizes and data structures used to save memory in preparation for mapping execution. Our evaluation shows that FlexRML’s adaptive bit size and data structure selection results in higher memory efficiency compared to conventional methods. When benchmarked against state-of-the-art RML processors, FlexRML consistently shows lower peak memory consumption across different datasets while delivering faster or comparable execution times.