Abstract. The Web of Linked Data forms a single, globally distributed dataspace. Due to the openness of this dataspace, it is not possible to know in advance all data sources that might be relevant for query answering. This openness poses a new challenge that is not addressed by traditional research on federated query processing. In this paper we present an approach to execute SPARQL queries over the Web of Linked Data. The main idea of our approach is to discover data that might be relevant for answering a query during the query execution itself. This discovery is driven by following RDF links between data sources based on URIs in the query and in partial results. The URIs are resolved over the HTTP protocol into RDF data which is continuously added to the queried dataset. This paper describes concepts and algorithms to implement our approach using an iterator-based pipeline. We introduce a formalization of the pipelining approach and show that classical iterators may cause blocking due to the latency of HTTP requests. To avoid blocking, we propose an extension of the iterator paradigm. The evaluation of our approach shows its strengths as well as the still existing challenges.

No abstract

Complex queries often contain common or similar subexpressions, either within a single query or among multiple queries submitted as a batch. If so, query execution time can be improved by evaluating a common subexpression once and reusing the result in multiple places. However, current query optimizers do not recognize and exploit similar subexpressions, even within the same query.We present an efficient, scalable, and principled solution to this long-standing optimization problem. We introduce a light-weight and effective mechanism to detect potential sharing opportunities among expressions. Candidate covering subexpressions are constructed and optimization is resumed to determine which, if any, such subexpressions to include in the final query plan. The chosen subexpression(s) are computed only once and the results are reused to answer other parts of queries. Our solution automatically applies to optimization of query batches, nested queries, and maintenance of multiple materialized views. It is the first comprehensive solution covering all aspects of the problem: detection, construction, and cost-based optimization. Experiments on Microsoft SQL Server show significant performance improvements with minimal overhead.

For many information domains there are numerous World Wide Web data sources. The sources vary both in their extension and their intension: They represent different real world entities with possible overlap and provide different attributes of these entities. Mediator-based information systems allow integrated access to such sources by providing a common schema against which the user can pose queries. Given a query, the mediator must determine which participating sources to access and how to integrate the incoming results.This article describes how to support mediators in their source selection and query planning process. We propose three new merge operators, which formalize the integration of multiple source responses. A completeness model describes the usefulness of a source to answer a query. The completeness measure incorporates both extensional value (called coverage) and intensional value (called density) of a source. We show how to determine the completeness of single sources and of combinations of sources under the new merge operators. Finally, we show how to use the measure for source selection and query planning.

Query execution over the Web of Linked Data has attracted much attention recently. A particularly interesting approach is link traversal based query execution which proposes to integrate the traversal of data links into the creation of query results. Hence -in contrast to traditional query execution paradigms-this does not assume a fixed set of relevant data sources beforehand; instead, the traversal process discovers data and data sources on the fly and, thus, enables applications to tap the full potential of the Web.While several authors have studied possibilities to implement the idea of link traversal based query execution and to optimize query execution in this context, no work exists that discusses theoretical foundations of the approach in general. Our paper fills this gap.We introduce a well-defined semantics for queries that may be executed using a link traversal based approach. Based on this semantics we formally analyze properties of such queries. In particular, we study the computability of queries as well as the implications of querying a potentially infinite Web of Linked Data. Our results show that query computation in general is not guaranteed to terminate and that for any given query it is undecidable whether the execution terminates. Furthermore, we define an abstract execution model that captures the integration of link traversal into the query execution process. Based on this model we prove the soundness and completeness of link traversal based query execution and analyze an existing implementation approach.

IntroductionMedical errors have an incidence of 9% and may lead to worse patient outcome. Teamwork training has the capacity to significantly reduce medical errors and therefore improve patient outcome. One common framework for teamwork training is crisis resource management, adapted from aviation and usually trained in simulation settings. Debriefing after simulation is thought to be crucial to learning teamwork-related concepts and behaviours but it remains unclear how best to debrief these aspects. Furthermore, teamwork-training sessions and studies examining education effects on undergraduates are rare. The study aims to evaluate the effects of two teamwork-focused debriefings on team performance after an extensive medical student teamwork training.Methods and analysesA prospective experimental study has been designed to compare a well-established three-phase debriefing method (gather–analyse–summarise; the GAS method) to a newly developed and more structured debriefing approach that extends the GAS method with TeamTAG (teamwork techniques analysis grid). TeamTAG is a cognitive aid listing preselected teamwork principles and descriptions of behavioural anchors that serve as observable patterns of teamwork and is supposed to help structure teamwork-focused debriefing. Both debriefing methods will be tested during an emergency room teamwork-training simulation comprising six emergency medicine cases faced by 35 final-year medical students in teams of five. Teams will be randomised into the two debriefing conditions. Team performance during simulation and the number of principles discussed during debriefing will be evaluated. Learning opportunities, helpfulness and feasibility will be rated by participants and instructors. Analyses will include descriptive, inferential and explorative statistics.Ethics and disseminationThe study protocol was approved by the institutional office for data protection and the ethics committee of Charité Medical School Berlin and registered under EA2/172/16. All students will participate voluntarily and will sign an informed consent after receiving written and oral information about the study. Results will be published.

No abstract

Set similarity joins, which compute pairs of similar sets, constitute an important operator primitive in a variety of applications, including applications that must process large amounts of data. To handle these data volumes, several distributed set similarity join algorithms have been proposed. Unfortunately, little is known about the relative performance, strengths and weaknesses of these techniques. Previous comparisons are limited to a small subset of relevant algorithms, and the large differences in the various test setups make it hard to draw overall conclusions. In this paper we survey ten recent, distributed set similarity join algorithms, all based on the MapReduce paradigm. We empirically compare the algorithms in a uniform test environment on twelve datasets that expose different characteristics and represent a broad range of applications. Our experiments yield a surprising result: All algorithms in our test fail to scale for at least one dataset and are sensitive to long sets, frequent set elements, low similarity thresholds, or a combination thereof. Interestingly, some algorithms even fail to handle the small datasets that can easily be processed in a non-distributed setting. Our analytic investigation of the algorithms pinpoints the reasons for the poor performance and targeted experiments confirm our analytic findings. Based on our investigation, we suggest directions for future research in the area.