Abstract:
Knowledge representation in RDF guarantees shared semantics and enables interoperability in data exchanges. Various approaches have been proposed for RDF knowledge graph construction, with declarative mapping languages emerging as the most reliable and reproducible solutions. However, not all information systems can understand and process data encoded as RDF. In these scenarios, to guarantee seamless communication there is a need for a further conversion of RDF graphs to one or more target data formats and models. Existing solutions for the declarative lifting of data to RDF are not able to effectively support knowledge conversion towards a generic output.
Based on an examination of existing mapping languages and processors for RDF knowledge graph construction, we define a reference workflow supporting a knowledge conversion process between different data representations.
The proposed workflow is validated by the mapping-template tool, an open-source implementation based on a popular template engine. The template-based mapping language enables the definition of mappings without requiring prior knowledge of RDF and provides flexibility for the target output. The tool is evaluated qualitatively, considering common challenges in the declarative specification of mappings, and quantitatively, considering performance and scalability.
This paper extends a previous version of this work by integrating a discussion of the proposed workflow considering the analysed state-of-the-art for knowledge graph construction, introducing the tool's direct support for the execution of RML mapping rules, and describing a more comprehensive qualitative and quantitative evaluation, also considering the results obtained by participating in the Knowledge Graph Construction Challenge 2024.
