Abstract

Despite the growing popularity of knowledge graphs for managing diverse data at large scale, users who wish to pose expressive queries against such graphs are often expected to know (i) how to formulate queries in a language such as SPARQL, and (ii) how entities of interest are described in the graph. In this paper we propose a language that relaxes these expectations; the language's operators are based on an interactive graph-based exploration that allows non-expert users to simultaneously navigate and query knowledge graphs; we compare the expressivity of this language with SPARQL. We then discuss an implementation of this language that we call RDF Explorer and discuss various desirable properties it has, such as avoiding interactions that lead to empty results. Through a user study over the Wikidata knowledgegraph, we show that users successfully complete more tasks with RDF Explorer than with the existing Wikidata Query Helper, while a usability questionnaire demonstrates that users generally prefer our tool and self-report lower levels of frustration and mental effort.