Abstract. Linked Data is at its core about the setting of links between resources. Links provide enriched semantics, pointers to extra information and enable the merging of data sets. However, as the amount of Linked Data has grown, there has been the need to automate the creation of links and such automated approaches can create low-quality links or unsuitable network structures. In particular, it is difficult to know whether the links introduced improve or diminish the quality of Linked Data. In this paper, we present LINK-QA, an extensible framework that allows for the assessment of Linked Data mappings using network metrics. We test five metrics using this framework on a set of known good and bad links generated by a common mapping system, and show the behaviour of those metrics.

Abstract. We present a technique for answering queries over RDF data through an evolutionary search algorithm, using fingerprinting and Bloom filters for rapid approximate evaluation of generated solutions. Our evolutionary approach has several advantages compared to traditional databasestyle query answering. First, the result quality increases monotonically and converges with each evolution, offering "anytime" behaviour with arbitrary trade-off between computation time and query results; in addition, the level of approximation can be tuned by varying the size of the Bloom filters. Secondly, through Bloom filter compression we can fit large graphs in main memory, reducing the need for disk I/O during query evaluation. Finally, since the individuals evolve independently, parallel execution is straightforward. We present our prototype that evaluates basic SPARQL queries over arbitrary RDF graphs and show initial results over large datasets.

Abstract. The Web of Data is increasingly becoming an important infrastructure for such diverse sectors as entertainment, government, ecommerce and science. As a result, the robustness of this Web of Data is now crucial. Prior studies show that the Web of Data is strongly dependent on a small number of central hubs, making it highly vulnerable to single points of failure. In this paper, we present concepts and algorithms to analyse and repair the brittleness of the Web of Data. We apply these on a substantial subset of it, the 2010 Billion Triple Challenge dataset. We first distinguish the physical structure of the Web of Data from its semantic structure. For both of these structures, we then calculate their robustness, taking betweenness centrality as a robustnessmeasure. To the best of our knowledge, this is the first time that such robustness-indicators have been calculated for the Web of Data. Finally, we determine which links should be added to the Web of Data in order to improve its robustness most effectively. We are able to determine such links by interpreting the question as a very large optimisation problem and deploying an evolutionary algorithm to solve this problem. We believe that with this work, we offer an effective method to analyse and improve the most important structure that the Semantic Web community has constructed to date.

The Linked Data cloud has grown to become the largest knowledge base ever constructed. Its size is now turning into a major bottleneck for many applications. In order to facilitate access to this structured information, this paper proposes an automatic sampling method targeted at maximizing answer coverage for applications using SPARQL querying. The approach presented in this paper is novel: no similar RDF sampling approach exist. Additionally, the concept of creating a sample aimed at maximizing SPARQL answer coverage, is unique. We empirically show that the relevance of triples for sampling (a semantic notion) is influenced by the topology of the graph (purely structural), and can be determined without prior knowledge of the queries. Experiments show a significantly higher recall of topology based sampling methods over random and naive baseline approaches (e.g. up to 90% for Open-BioMed at a sample size of 6%).

Abstract. Here, we describe the CEDAR dataset, a five-star Linked Open Data representation of the Dutch historical censuses. These were conducted in the Netherlands once every 10 years from 1795 to 1971. We produce a linked dataset from a digitized sample of 2,288 tables. It contains more than 6.8 million statistical observations about the demography, labour and housing of Dutch society in the 18th, 19th and 20th centuries. The dataset is modeled using the RDF Data Cube, Open Annotation, and PROV vocabularies. These are used to represent the multidimensionality of the data, to express rules of data harmonization, and to keep track of the provenance of all data points and their transformations, respectively. We link observations within the dataset to well known standard classification systems in social history, such as the Historical International Standard Classification of Occupations (HISCO) and the Amsterdamse Code (AC). The three contributions of the dataset are (1) an easier access to integrated census data for historical researchers; (2) richer connections to related Linked Data resources; and (3) novel concept schemes of historical relevance, like classifications of historical religions and historical house types.