The Virtual Knowledge Graph System Ontop GuohuiXiao Free University of Bozen-Bolzano
Italy
Ontopic s.r.l
Bolzano Italy
DavideLanti Free University of Bozen-Bolzano
Italy
RomanKontchakov Birkbeck University of London
UK
SarahKomla-Ebri Ontopic s.r.l
Bolzano Italy
ElemGüzel-Kalaycı elem.guezelkalayci@v2c2.at Virtual Vehicle Research GmbH
Graz Austria
LinfangDing Free University of Bozen-Bolzano
Italy
JulienCorman Free University of Bozen-Bolzano
Italy
BenjaminCogrel Ontopic s.r.l
Bolzano Italy
DiegoCalvanese diego.calvanese@umu.se Free University of Bozen-Bolzano
Italy
Ontopic s.r.l
Bolzano Italy
Umeå University
Sweden
ElenaBotoeva e.botoeva@imperial.ac.uk Imperial College London
UK
The Virtual Knowledge Graph System Ontop 0120E0C6A837D777C3AA0F71AE8F5E28 GROBID - A machine learning software for extracting information from scholarly documents

The full paper is published at the Resource Track of ISWC 2020 [14] VKG. The Virtual Knowledge Graph (VKG) approach, also known in the literature as Ontology-Based Data Access (OBDA) [8,12], has become a popular paradigm for accessing and integrating data sources [13]. In such approach, the data sources, which are normally relational databases, are virtualized through a mapping and an ontology, and presented as a unified knowledge graph, which can be queried by end-users through a vocabulary they are familiar with. At query time, a VKG system translates user queries over the ontology to SQL queries over the database. This approach frees end-users from the low-level details of data organization, so that they can concentrate on their high-level tasks. As it is gaining more importance, the VKG paradigm has been implemented in several systems [1,2,9,11] and adopted in a wide range of use cases. Here, we present the latest major release, Ontop v4, of a popular VKG system. Ontop v1. The development of Ontop has spanned the past decade. Developing such a system is highly non-trivial and requires both a theoretical investigation of the semantics and strong engineering efforts to implement all the required features. Ontop started in 2009, only one year after the first version of SPARQL had been standardized, while OWL 2 QL [6] and R2RML [4] appeared 3 years later, in 2012. At that time, the VKG research focused on union of conjunctive queries (UCQs) as a query language. With this target, Ontop v1 relied on non-recursive Datalog as its core data structure [10] because it perfectly fit the UCQ-based setting. The development of Ontop was boosted by the EU FP7 project Optique (2013-2016), during which the compliance with the relevant W3C recommendations became a priority, and significant progress was made in this direction. The last release of Ontop v1 was v1.18 in 2016 [1]. New challenges. A natural requirement that emerged during the Optique project were aggregates introduced in SPARQL 1.1 [5]. The Ontop development team spent a major effort, internally called Ontop v2, on implementing this query language feature. However, it became exceedingly clear that the Datalog representation was not well suited for this implementation. Some prototypes of Ontop v2 were used in the Optique project for internal purposes, but never reached the level of a public release. During this development, as Ontop moved towards supporting the W3C recommendations for SPARQL and R2RML, we have identified the following new challenges:

-In contrast to the usual DL encoding with unary and binary predicates for classes and properties, in SPARQL triple pattern variables can occur in positions of class and property names, which means that there are effectively only two 'predicates': triple for triples in the RDF dataset default graph, and quad for named graphs. -More importantly, SPARQL is based on a rich algebra, which goes beyond the expressivity of CQs. Non-monotonic features like optional and minus, cardinality-sensitive query modifiers (distinct) and aggregation (group by with functions such as sum, avg, count) are difficult to model even in extensions of Datalog. -Even without SPARQL aggregation, cardinalities have to be treated carefully: the SQL queries in a mapping produce bags of tuples, but their induced RDF graphs contain no duplicates and thus are sets of triples; however, when a SPARQL query is evaluated, it results in a bag of solutions mappings. These challenges turned out to be difficult to tackle in the Datalog setting.

Ontop v4. To address the challenges posed by aggregation, and others that had emerged in the meantime, we started to investigate an alternative core data structure. The outcome has been what we call intermediate query (IQ), an algebra-based data structure that unifies both SPARQL and relational algebra. Using IQ, we reimplemented most of the Ontop code base. After two beta releases in 2017 and 2018, we released the stable version of Ontop v3 in 2019. Following Ontop v3, the development focussed on improving compliance and adding several major features. In particular, aggregates have now been supported since Ontop v4-beta-1, released in late 2019. The stable version of Ontop v4 was released in July 20201 . The documentation is provided at the official website2 .

Evaluation. Ontop v4 has greatly improved its compliance with relevant W3C recommendations and provides good performance in query answering. It supports almost all the features of SPARQL 1.1, R2RML, OWL 2 QL, and SPARQL entailment regime, and the SPARQL 1.1 HTTP Protocol. In particular, in Table 1, we present a summary of Ontop v4 compliance with SPARQL 1.1, where rows correspond to sections of the WC3 recommendation. Most of the features are supported, but some are unsupported or only partially supported. Note that most of the missing SPARQL functions (Section 17.4) are not so challenging to implement but require a considerable engineering effort to carefully define their translations into SQL. We will continue the process of implementing them gradually and track the progress in a dedicated issue 3 . Recently, two indepen-

Table 1 .SPARQL Compliance: unsupported features are crossed out. When considering all the perspectives, like usability, completeness, and soundness, Ontop clearly stands out among the open-source systems. Community and Adoption. Ontop is the result of an active developer community. It has been downloaded more than 30K times from Sourceforge. In addition to the research groups, Ontop is also backed by a commercial company, Ontopic s.r.l., born in April 2019. Ontop has been adopted in many academic and industrial use cases. However, due to its liberal Apache 2 license, it is essentially impossible to obtain a complete picture of all use cases and adoptions. Nevertheless, a few significant use cases have been summarized in a recent survey paper[13]. Finally, we mention two recent commercial deployments of Ontop: UNiCS (http://unics.cloud/) is an open data platform for research and innovation, and ODH-VKG (https://sparql.opendatahub.bz.it/) is a project publishing South Tyrolean tourism data as a Knowledge Graph.Section in SPARQL 1.1 [5]FeaturesCoverage5-7. Graph Patterns, etc.BGP, FILTER, OPTIONAL, UNION4/48. NegationMINUS, FILTER [NOT] EXISTS1/29. Property PathsPredicatePath, InversePath, ZeroOrMorePath, . . .010. AssignmentBIND, VALUES2/211. AggregatesCOUNT, SUM, MIN, MAX, AVG, GROUP_CONCAT, SAMPLE6/612. SubqueriesSubqueries1/113. RDF DatasetGRAPH, FROM [NAMED]1/214. Basic Federated QuerySERVICE015. Solution Seqs. & Mods.ORDER BY, SELECT, DISTINCT, REDUCED, OFFSET, LIMIT6/616. Query FormsSELECT, CONSTRUCT, ASK, DESCRIBE4/417.4.1. Functional FormsBOUND, IF, COALESCE, EXISTS, NOT EXISTS,6/11||, &&, =, sameTerm, IN, NOT IN17.4.2. Fns. on RDF TermsisIRI, isBlank, isLiteral, isNumeric, str, lang,9/13datatype, IRI, BNODE, STRDT, STRLANG, UUID, STRUUID17.4.3. Fns. on StringsSTRLEN, SUBSTR, UCASE, LCASE, STRSTARTS, STRENDS,14/14CONTAINS, STRBEFORE, STRAFTER, ENCODE_FOR_URI,CONCAT, langMatches, REGEX, REPLACE17.4.4. Fns. on Numericsabs, round, ceil, floor, RAND5/517.4.5. Fns. on Dates&Times now, year, month, day, hours,8/9minutes, seconds, timezone, tz17.4.6. Hash FunctionsMD5, SHA1, SHA256, SHA384, SHA5125/517.5. XPath Constructor Fns. casting017.6. Extensible Value Testing user defined functions0dent evaluations [3,7] of VKG systems have confirmed the robust performance ofOntop.
https://github.com/ontop/ontop https://ontop-vkg.org/ https://github.com/ontop/ontop/issues/346
Ontop: answering SPARQL queries over relational databases DCalvanese BCogrel SKomla-Ebri RKontchakov DLanti MRezk MRodriguez-Muro GXiao SWJ 8 3 2017 The MASTRO system for ontologybased data access DCalvanese GDe Giacomo DLembo MLenzerini APoggi MRodriguez-Muro RRosati MRuzzi DFSavo SWJ 2 1 2011 Using Berlin SPARQL benchmark to evaluate relational database virtual SPARQL endpoints MChaloupka MNecasky 2020 Submitted to SWJ R2RML: RDB to RDF mapping language SDas SSundara RCyganiak W3C recommendation W3C 2012 SPARQL 1.1 query language SHarris ASeaborne EPrud Hommeaux W3C recommendation W3C 2013 OWL 2 Web Ontology Language: Profiles BMotik BGrau IHorrocks ZWu AFokoue CLutz W3C Recommendation
W3C
 2012 Comparing query answering in OBDA tools over W3C-compliant specifications MNamici GDeGiacomo Proc. DL DL 2018 2211 Linking data to ontologies APoggi DLembo DCalvanese GDe Giacomo MLenzerini RRosati J. Data Sem 10 2008 Formalisation and experiences of R2RML-based SPARQL to SQL query translation using morph FPriyatna ÓCorcho JFSequeda WWW 2014 Efficient SPARQL-to-SQL with R2RML mappings MRodriguez-Muro MRezk J. Web Sem 33 2015 Ontology-based data access using views JFSequeda MArenas DPMiranker Proc. RR RR Springer 2012 7497 Ontology-based data access: A survey GXiao DCalvanese RKontchakov DLembo APoggi RRosati MZakharyaschev Proc. IJCAI IJCAI 2018 Virtual knowledge graphs: An overview of systems and use cases GXiao LDing BCogrel DCalvanese Data Intelligence 1 2019 The virtual knowledge graph system ontop GXiao DLanti RKontchakov SKomla-Ebri EGüzel-Kalayci LDing JCorman BCogrel DCalvanese EBotoeva ISWC 2020