SPARQL and SQL: technical aspects and synergy
- 1. SPARQL and SQL technical aspects and synergy Relational model flat model, tabular form, typing is implicit (by convention, column definition) JOIN functions used to combine information from tables use of foreign keys to add semantics and graph-like structures each table has many columns = many attributes of objects RDF model may be viewed as flat but because it is a graph can also be traversed hierarchically or arbitrarily explicit typing is common (by namespace definition) URIs instead of closed database table names explicit relationships via predicates (triple format: subj, pred, obj) relationships and references are the key point SPARQL designed to query collections of triples and to easily traverse relationships syntax vaguely resembles SQL (SELECT, WHERE, etc) it matches graph patterns SPARQL and SQL SPARQL can be translated to SQL
- 2. SPARQL “under the hood” Works with popular web and web services protocols: HTTP and SOAP XML results format (easy to transform, XSLT, XQuery) A SPARQL query consists of three parts: Pattern matching (OPTIONAL, UNION, FILTER) Solution modifiers (PROJECTION, DISTINCT, OFFSET, LIMIT) Output part CONSTRUCT of new triples BGP (Basic Graph Pattern) built-in mechanism Can work in conjunction with RDFS ’s taxonomic reasoning (respects RDFS type checking and subsumption thus providing inference) Inference system in RDFS (and through SPARQL) supports existential, sub-property, sub-class, and typing (implicit and explicit) inferences
- 3. SPARQL applications, engines, endpoints RDF engines (SPARQL to interrogate them) Oracle (will support SPARQL syntax), Allegrograph, OpenLink Virtuoso, ARQ/Joseki (HP), Boca (IBM), Rasqal for Redland, SWI-Prolog, Sesame, D2R Server Notable SPARQL users & applications (R&D and commercial) Garlik: SPARQL to built reports on people’s online credit reports, 500-2000 SPARQL queries to build a report, 1-2 seconds processing time, 8 big knowledge bases (>2bn triples), XML format) JSpace: an extension to mSpace, UI driven SPARQL firing, digital music domain POPS: expertise location service for NASA, federates 4 diverse data sources, 4.5M triples, pilot study for the Office of the Chief Engineer BIANCA: network asset management service, integrated view of applications, servers, networks and changes and their relations, deployed at NASA HCLS (Health Care and Life Sciences) interest group at W3C, 60 organisations members, Google maps based interface for Allen Brain Atlas, 20k genes, 400k images. Further resources and point of info: http://esw.w3.org/topic/SparqlImplementations
- 4. SPARQL and the Semantic Web - technical Relation of SPARQL to rules for the Semantic Web Rules: several proposals for rule systems usable on top of RDF/RDF (SWRL, N3, TRIPLE, SWI-Prolog, WSML, etc) Only a handful through W3C standardisation (SWRL) and already working on RIF (for interchanging rules) SPARQL is a query language (primarily) but can be used as a rule language (translation work underway for SPARQL to LP (Datalog) ) Example the RDFS entailment rule that can be expressed with SPARQL CONSTRUCT clause RIF charter states that it will be compatible with SPARQL
- 5. SPARQL and SQL technical aspects and synergy SQL Great for finding data from tabular representations, can get complex when many tables are involved in a given query SPARQL Good pattern matching paradigm, especially when relationships have to be used to answer a query Also add here: work on BGP for translating SPARQL to SQL (technical paper) Federated SPARQL endpoints and SPARQL validators, query resolvers, SPARQL APIs Use of SPARQL CONSTRUCTS as opposed to rules and inference engines Argument for SPARQL overarching SQL schemata: reduces computational workload with having to describe complex JOIN SQL queries; common namespaces and a match to underlying SQL schemata is the only workload Syndication and verticals to agree on common namespaces? Abstract layer from which to access merchant specific DBs Check sylvia’s slides from WSRI seminar – scalability of SPARQL and use cases – lack of SUM function, etc. – check also my slides Some examples of SPARQL at work? From the recent NaTii experiment and from e-response - show them data provenance and typing examples (use of namespaces)