MAPLE: metadata-driven orchestration of ontology matching

Bringing together disparate datasets on the Semantic Web clearly benefits from ontology matching. Systematic evaluation campaigns have focused on performance, efficiency, scalability and, more recently, human involvement. No less important is the recognition of the differences between datasets, for example in terms of modelling languages, lexicalisation and structure, which enables the selection and configuration of appropriate techniques and support resources. Following the Semantic Web vision of machines dialoguing to solve problems, we propose MAPLE, a framework that semi-automatically orchestrates an alignment plan using metadata about the matched datasets and other available resources. The framework prescribes a metadata profile that combines established vocabularies such as VoID, DCAT, Dublin Core and LIME, making it possible to use metadata accompanying self-describing datasets or published in catalogues. We discuss the integration of the framework into the collaborative knowledge development environment VocBench 3, as well as compatible matching systems.