Human-Friendly RDF Graph Construction: Which One Do You Chose?

Knowledge Graphs (KGs) are a powerful mechanism to structure and organize data on the Web. RDF KGs are usually constructed by declaring a set of mapping rules, specified according to the grammar of a mapping language (e.g., RML), that relates the input data sources to a domain vocabulary. However, the verbosity and (manual) definition of these rules affect their global adoption. Several user-friendly serializations for different mapping languages were proposed to facilitate users with the definition of such rules, e.g., YARRRML, SMS2, XRM, or ShExML. Still, most of them do not cover all features of the mapping languages for RDF graph construction (e.g., constructing RDF-star), or they lack tooling support. In this paper, (i) we present a set of updates over the YARRRML serialisation to empower it with the latest necessities for constructing RDF graphs; (ii) we implement these new features in a new open-source translator, Yatter, currently used in different real-use cases and international projects; and (iii) we qualitatively compare our proposal against similar state-of-the-art serialisations, and their associated translators over a set of conformance test cases. Our proposal advances the declarative construction of RDF graphs and supports users in choosing an appropriate serialisation and translator for their use cases.