In this work , we propose SPARQLGX: an implementation of a distributed RDF datastore based on Apache Spark. SPARQLGX is designed to leverage existing Hadoop infrastructures for evaluating SPARQL queries efficiently. SPARQLGX relies on an automated translation of SPARQL queries into optimized executable Spark code. We show that SPARQLGX makes it possible to evaluate SPARQL queries on billions of triples distributed across multiple nodes, while providing attractive performance figures. We report on experiments which show how SPARQLGX compares to state-of-the-art implementations and we show that our approach scales better than other systems in terms of supported dataset size. With its simple design, SPARQLGX represents an interesting alternative in several scenarios.

The increasing number of available datasets gives opportunities to build large and complex applications which aggregate results coming from several sources. These emerging use cases require new systems where the combinations of heterogeneous sources are both allowed and efficient. To tackle these challenges, we built a system offering a simple high-level set of primitives – called HAP – to easily describe processing chains. These descriptions are then compiled into optimized SQL queries executed on Hive.

In this work , we provide a new perspective on distributed sparql evaluators, based on a multi-criteria ranking obtained through extensive experiments. Specifically, we propose a set of five principal features which we use to rank evaluators. Each system exhibits a particular combination of these features. Similarly, the various requirements of practical use cases can also be decomposed in terms of these features. Our suggested set of features provides a more comprehensive description of the behavior of a distributed evaluator when compared to traditional performance metrics. We show how it helps in more accurately evaluating to which extent a given system is appropriate for a given use case. For this purpose, we systematically benchmarked a panel of 10 state-of-the-art implementations. We ranked them using this reading grid to pinpoint the advantages and limitations of current sparql evaluation systems.

One aim of the RDF data model, as standardized by the W3C, is to facilitate the evolution of data over time without requiring all the data consumers to be changed. To this end, one of the latest addition to the SPARQL standard query language is an update language for RDF graphs. The research on efficient and scalable SPARQL evaluation methods increasingly relies on standardized methodologies for benchmarking and comparing systems. However, current RDF benchmarks do not support graphs updates. We propose SPARUB: a benchmark for the SPARQL UPDATE language on RDF graphs . The aim of SPARUB is not to be yet another RDF benchmark. Instead it provides the mean to automatically extend and improve existing RDF benchmarks along a new dimension of data updates, while preserving their structure and query scenarios.

SPARQL is the w3c standard query language for querying data expressed in the Resource Description Framework (RDF). There exists a variety of SPARQL evaluation schemes and, in many of them, estimating the cardinality of intermediate results is key for performance, especially when the computation is distributed and the datasets very large. For example it helps in choosing join orders that minimize the size of intermediate subquery results.

In this context , we propose a new cardinality estimation based on statistics about the data. Our cardinality estimation is a worst-case analysis tailored for SPARQL and capable of taking advantage of the implicit schema often present in RDF datasets (e.g. functional dependencies). This implicit schema is captured by statistics therefore our method does not need for the schema to be explicit or perfect (our system performs well even if there are a few “violations” of these implicit dependencies). We implemented our cardinality estimation and used it to optimize the evaluation of SPARQL queries: equipped with our cardinality estimation, the query evaluator performs better against most queries (sometimes by an order of magnitude) and is only ever slightly slower.

In this work , , we propose a new algebra, $\mu$-algebra, inspired by works on the relational algebra, SQL and NoSQL languages (especially SPARQL) along with a prototype implementation of a SPARQL optimizer based on this algebra1. Our algebra has the following properties: (1) It subsumes the SPARQL Algebra (under the set semantics) with a more general recursion. (2) SPARQL with Property Paths can be e ciently translated to this algebra. (3) We have a type system and rewriting rules for terms of this algebra that allow optimization, notably of terms involving recursion. We illustrate the differences and the benefits of our approach on recursive query optimization. While a generic approach often comes at the cost of performance, we experimentally show that this approach actually leads to more efficient evaluation of queries with Property Paths. We also show that our approach produces Query Execution Plans (QEP) that are not considered by other existing approaches.

Data structured in the Resource Description Framework (RDF) are increasingly available in large volumes. This leads to a major need and research interest in novel methods for query analysis and compilation for making the most of RDF data extraction. SPARQL is the widely used and well supported standard query language for RDF data. In parallel to query language evolutions, schema languages for expressing constraints on RDF datasets also evolve. Shape Expressions (ShEx) are increasingly used to validate RDF data, and to communicate expected graph patterns. Schemas in general are important for static analysis tasks such as query optimisation and containment.

In this work , , we investigate the means and methodologies for SPARQL query static analysis in the presence of ShEx schema constraints. Our contribution consists in considering the problem of SPARQL query containment in the presence of ShEx constraints. We propose a sound and complete procedure for the problem of containment with ShEx, considering several SPARQL fragments. Particularly our procedure considers OPTIONAL query patterns, that turns out to be an important feature to be studied with schemas. We provide complexity bounds for the containment problem with respect to the language fragments considered. We also propose alternative method for SPARQL query containment with ShEx by reduction into First Order Logic satisfiability, which allows for considering SPARQL fragment extension in comparison to the first method. This is the first work addressing SPARQL query containment in the presence of ShEx constraints.

ShEx (Shape Expressions) is a language for expressing constraints on RDF graphs. In this work , , , we optimize the evaluation of conjunctive SPARQL queries, on RDF graphs, by taking advantage of ShEx constraints. Our optimization is based on computing and assigning ranks to query triple patterns, dictating their order of execution. The presence of intermediate joins between the query triple patterns is the reason why ordering is important in increasing efficiency. We first define a set of well formed ShEx schemas, that possess interesting characteristics for SPARQL query optimization. We then define our optimization method by exploiting information extracted from a ShEx schema. We finally report on evaluation results performed showing the advantages of applying our optimization on the top of an existing state-of-the-art query evaluation system.

In this work , we study the problem of enumerating the satisfying valuations of a circuit while bounding the delay, i.e., the time needed to compute each successive valuation. We focus on the class of structured d-DNNF circuits originally introduced in knowledge compilation, a sub-area of artificial intelligence. We propose an algorithm for these circuits that enumerates valuations with linear preprocessing and delay linear in the Hamming weight of each valuation. Moreover, valuations of constant Hamming weight can be enumerated with linear preprocessing and constant delay. Our results yield a framework for efficient enumeration that applies to all problems whose solutions can be compiled to structured d-DNNFs. In particular, we use it to recapture classical results in database theory, for factorized database representations and for MSO evaluation. This gives an independent proof of constant-delay enumeration for MSO formulae with first-order free variables on bounded-treewidth structures.

Although XQuery is a statically typed, functional query language for XML data, some of its features such as upward and horizontal XPath axes are typed imprecisely. The main reason is that while the XQuery data model allows us to navigate upwards and between siblings from a given XML node, the type model, e.g., regular tree types, can describe only the subtree structure of the given node. Recently, Castagna et al. (2015) and Genevès and Gesbert (2015) independently propose a precise forward type inference system for XQuery using an extended type language that can describe not only a given XML node but also its context. In this work , as a complementary method to forward type inference systems, we propose a novel backward type inference system for XQuery, using the type language proposed by Genevès and Gesbert (2015). Our backward type inference system provides an exact typing result for XPath axes and a sound typing result for XQuery expressions.

In this work , we show how the analysis of very large amounts of drug prescription data make it possible to detect, on the day of hospital admission, patients at risk of developing complications during their hospital stay. We explore, for the first time, to which extent volume and variety of big prescription data help in constructing predictive models for the automatic detection of at-risk profiles. Our methodology is designed to validate our claims that: (1) drug prescription data on the day of admission contain rich information about the patient's situation and perspectives of evolution, and (2) the various perspectives of big medical data (such as veracity, volume, variety) help in extracting this information. We build binary classification models to identify at-risk patient profiles. We use a distributed architecture to ensure scalability of model construction with large volumes of medical records and clinical data. We report on practical experiments with real data of millions of patients and hundreds of hospitals. We demonstrate how the fine-grained analysis of such big data can improve the detection of at-risk patients, making it possible to construct more accurate predictive models that significantly benefit from volume and variety, while satisfying important criteria to be deployed in hospitals.