Search

SPARQL query services that balance processing between clients and servers become more and more essential to handle the increasing load for open and decentralized knowledge graphs on the Web. To this end, Linked Data Fragments (LDF) have introduced a foundational framework that has sparked research exploring a spectrum of potential Web querying interfaces in between server-side query processing via SPARQL endpoints and client-side query processing of data dumps. Current proposals in between typically suffer from imbalanced load on either the client or the server. In this paper, to the best of our knowledge, we present the first work that combines both client-side and server-side query optimization techniques in a truly dynamic fashion: we introduce WiseKG, a system that employs a cost model that dynamically delegates the load between servers and clients by combining client-side processing of shipped partitions with efficient server-side processing of star-shaped sub-queries, based on current server workload and client capabilities. Our experiments show that WiseKG significantly outperforms state-of-the-art solutions in terms of average total query execution time per client, while at the same time decreasing network traffic and increasing server-side availability.

A multitude of Linked Data Fragments (LDF) server interfaces have been proposed to expose Knowledge Graphs (KGs) on the Web. Each interface leads to different trade-offs when clients execute queries over them, such as how query execution effort is distributed between server and client. There is however no single silver bullet that works best everywhere. Each of these interfaces has diverse characteristics that vary the performance based on server load, client resources, and network bandwidth. Currently, publishers can only pick one of these interfaces to expose their KGs on the Web. However, in some cases, multiple interfaces may be suitable for the publisher, and these may even vary over time based on the aforementioned factors. As such, we propose a hybrid LDF interface that can expose multiple interfaces based on a server-side cost model. Additionally, we sketch a negotiation protocol through which clients can determine desirable interfaces during query planning using a client-side cost model. In this paper, we lay out the high-level ideas behind this hybrid framework, and we explain our future steps regarding implementation and evaluation. As such, our work provides a basis for exploiting the trade-offs that exist between different LDF interfaces for optimally exposing KGs on the Web.

While Linked Data (LD) provides standards for publishing (RDF) and (SPARQL) querying Knowledge Graphs (KGs) on the Web, serving, accessing and processing such open, decentralized KGs is often practically impossible, as query timeouts on publicly available SPARQL endpoints show. Alternative solutions such as Triple Pattern Fragments (TPF) attempt to tackle the problem of availability by pushing query processing workload to the client side, but suffer from unnecessary transfer of irrelevant data on complex queries with large intermediate results. In this paper we present smart-KG, a novel approach to share the load between servers and clients, while significantly reducing data transfer volume, by combining TPF with shipping compressed KG partitions. Our evaluations show that smart-KG outperforms state-of-the-art client-side solutions and increases server-side availability towards more cost-effective and balanced hosting of open and decentralized KGs.

Smart city infrastructures such as transportation and energy networks are evolving into so-called cyber physical social systems (CPSSs), which collect and leverage citizens’ data in order to adapt services to citizens’ needs. The privacy implications of such systems are, however, significant and need to be addressed. Current systems either try to escape the privacy challenge via anonymization or use very rigid, hard-coded workflows that have been agreed with a data protection authority. In the case of the latter, there is a severe impact on data quality and richness, whereas in the former, only these hard-coded flows are permitted resulting in diminished functionality and potential. We address these limitations via user modeling in terms of investigating how to model and semantically represent user consent, preferences, and data usage policies that will guide the processing of said data in the data lake. Data protection is a horizontal field and consequently very wide. Therefore, we focus on a concrete setting where we extend the domain-agnostic SPECIAL policy language for a smart mobility use case supplied by Vienna’s largest utility provider. To that end, (1) we create an extension of SPECIAL in terms of a core CPSS vocabulary that lowers the semantic gap between the domain agnostic terms of SPECIAL and the vocabulary of the use case; (2) we propose a workflow that supports defining domain-specific vocabularies for complex CPSSs; and (3) show that these two contributions allow successfully achieving the goals of our setting.

The European General Data Protection Regulation (GDPR) brings new challenges for companies, who must demonstrate that their systems and business processes comply with usage constraints specified by data subjects. However, due to the lack of standards, tools, and best practices, many organizations struggle to adapt their infrastructure and processes to ensure and demonstrate that all data processing is in compliance with users' given consent. The SPECIAL EU H2020 project has developed vocabularies that can formally describe data subjects' given consent as well as methods that use this description to automatically determine whether processing of the data according to a given policy is compliant with the given consent. Whereas this makes it possible to determine whether processing was compliant or not, integration of the approach into existing line of business applications and ex-ante compliance checking remains an open challenge. In this short paper, we demonstrate how the SPECIAL consent and compliance framework can be integrated into Linked Widgets, a mashup platform, in order to support privacy-aware ad-hoc integration of personal data. The resulting environment makes it possible to create data integration and processing workflows out of components that inherently respect usage policies of the data that is being processed and are able to demonstrate compliance. We provide an overview of the necessary meta data and orchestration towards a privacy-aware linked data mashup platform that automatically respects subjects' given consents. The evaluation results show the potential of our approach for ex-ante usage policy compliance checking within the Linked Widgets Platforms and beyond.

In the last decade, the Resource Description Framework (RDF) has become the de-facto standard for publishing semantic data on the Web. This steady adoption has led to a significant increase in the number and volume of available RDF datasets, exceeding the capabilities of traditional RDF stores. This scenario has introduced severe big semantic data challenges when it comes to managing and querying RDF data at Web scale. Despite the existence of various off-the-shelf Big Data platforms, processing RDF in a distributed environment remains a significant challenge. In this position paper, based on an indepth analysis of the state of the art, we propose to manage large RDF datasets in Flink, a well-known scalable distributed Big Data processing framework. Our approach, which we refer to as FLINKer extends the native graph abstraction of Flink, called Gelly, with RDF graph and SPARQL query processing capabilities.

In this paper, we investigate the problem of recognizing entities which are targeted by text sentiment in Arabic tweets. To do so, we train a bidirectional LSTM deep neural network with conditional random fields as a classification layer on top of the network to discover the features of this specific set of entities and extract them from Arabic tweets. We’ve evaluated the network performance against a baseline method which makes use of a regular named entity recognizer and a sentiment analyzer. The deep neural network has shown a noticeable advantage in extracting sentiment target entities from Arabic tweets.

Online Communities for Question Answering (CQA) such as Quora and Stack Overflow face the challenge of providing sufficient answers for the questions asked by users. The exponential growing rate of the unanswered questions compromises the effectiveness of the CQA frameworks as knowledge sharing platforms. The main reason for this issue is the inefficient routing of the questions to the potential answerers, who are the field experts and interested users. This paper proposes the deep-learning-based technique QR-DSSM to increase the accuracy of the question routing process. This technique uses deep semantic similarity model (DSSM) to extract semantic similarity features using deep neural networks and use the features to rank users' profiles. QR-DSSM maps the asked questions and the profiles of the users into a latent semantic space where the ability to answer is measured using the cosine similarity between the questions and the profiles of the users. QR-DSSM experiments outperformed the baseline models such as LDA, SVM, and Rank-SVM techniques and achieved an MRR score of 0.1737.

Online Communities for Question Answering (CQA) such as Quora and Stack Overflow face the challenge of providing high quality answers to the questions asked by their users. Although CQA frameworks receive new questions in a linear rate, the rate of the unanswered questions increases in an exponential way. This variation eventually compromise effectiveness of the CQA frameworks as knowledge sharing platforms. The main cause for this challenge is the improper routing of questions to the potential answerers, field experts or interested users. The proposed technique QR-DSSM uses deep semantic similarity model (DSSM) to extract semantic similarity features using deep neural networks. The extracted semantic features are used to rank the profiles of the answerers by their relevance the routed question. QR-DSSM maps the asked questions and the profiles of the users into a latent semantic space where the relevance is measured using cosine similarity between the two; questions and users’ profiles. QR-DSSM achieved MRR score of 0.1737. QR-DSSM outperformed the baseline models such as query likelihood language model (QLLM), Latent Dirichlet Allocation (LDA), SVM classification technique and RankingSVM learning to rank technique.