DARQL: Deep Analysis of SPARQL Queries

Abstract

In this demonstration, we showcase DARQL, the first tool for deep, large-scale analysis of SPARQL queries. We have harvested a large corpus of query logs with different lineage and sizes, from DBPedia to BioPortal and Wikidata, whose total number of queries amounts to 180M. We ran a wide range of analyses on the corpus, spanning from simple tasks (keyword counts, triple counts, operator distributions), moderately deep tasks (projection test, query classification), and deep analysis (shape analysis, well-designedness, weakly well-designedness, hypertreewidth, and fractional edge cover). The key goal of our demonstration is to let the users dive into the SPARQL query logs of our corpus and let them discover the inherent characteristics of the queries. The entire corpus of SPARQL queries is stored in a DBMS. The tool has a GUI that allows users to ask sophisticated analytical queries on the SPARQL logs. These analytical queries can both be directly written in SQL or composed by a visual query builder tool. The results of the analytical queries are represented both textually (as SPARQL queries) and visually. The DBMS performs the searches within the corpus quite efficiently. To the best of our knowledge, this is the first demonstration of this kind on such a large corpus and with such a number of varied tests.