Assigning semantic labels to table columns and identifying relations between columns pose significant challenges in data management. Automatic column annotation has been widely treated as classification, with recent works using language models trained on annotated tables with type and property labels. While these language models have effectively modeled individual tables, they often overlook the underlying graph structure of the label space, where constraints can exist between certain types and properties within and across tables. To fill this gap, we propose RODEO, a two-tower architecture that integrates a language model and a graph neural network (GNN) to model the table and semantic labels, respectively. We reformulate column annotation tasks from classification to matching problems, where column and column-pair embeddings are aligned with embeddings that represent their corresponding semantic types (nodes) and properties (edges) within the graph. These embeddings, derived from the language model and GNN, are co-trained end-to-end using triplet loss with an online negative mining strategy. The training process brings semantically related columns and labels closer in the embedding space by minimizing their distances. Our approach, evaluated on publicly available benchmark datasets, outperforms state-of-the-art methods in both column type and column property annotation, highlighting that modeling label constraints through the graph significantly improves overall performance. Ablation studies on the triplet loss and GNN show the robustness of our framework's training procedure.

Data-driven innovation has recently changed the mindset in data sharing from centralized architectures and monolithic data exploitation by data providers (data platforms) to decentralized architectures and different data sharing options among all involved participants (data ecosystems). Data sharing is further strengthened through the establishment of several legal frameworks (e.g., European Strategy for Data, Data Act, Data Governance Act) and the emerging initiatives that provide the means to build data ecosystems, which is evident in the formulated communities, established use cases, and the technical solutions. However, the data ecosystems have not been thoroughly studied so far. The differences between the various data ecosystems are not clear, making it hard to choose the most suitable for each use case, negatively impacting their adoption. Since the domain is growing fast, a review of the state-of-the-art data ecosystem initiatives is needed to analyze what each initiative offers, identify collaboration prospects, and highlight features for improvement and open research topics. In this paper, we review the state-of-the-art data ecosystem initiatives, describe their innovative aspects, compare their technical and business features, and identify open research challenges. We aim to assist practitioners in choosing the most suitable data ecosystem for their use cases and scientists to explore emerging research opportunities. Furthermore, we will provide a framework that outlines the key criteria for evaluating these initiatives, ensuring that stakeholders can make informed decisions based on their specific needs and objectives. By synthesizing our findings, we hope to foster a deeper understanding of the evolving landscape of data ecosystems and encourage further advancements in this critical field.

Knowledge graphs (KGs) play a crucial role in the integration and organization of heterogeneous data and knowledge, enabling advanced data analytics and decision-making across various industries. This vision paper addresses critical challenges in managing KGs, emphasizing their relevance in integrating information from disparate sources. We propose the concept of knowledge graph ecosystems and life cycles to systematically manage tasks, e.g., data integration, standardization, continuous updates, efficient querying, and provenance tracking. By adopting our approach, organizations can enhance the accuracy, consistency, and reliability of KGs, thus improving knowledge management, enabling the extraction of valuable insights, and ensuring transparency and accountability.

SHACL shapes describe the constraints of RDF subgraphs which are constructed from heterogeneous data, such as RDBs, JSONs, XMLs, etc. These heterogeneous data often already have constraints defined in their schemas, e.g., JSON Schema for JSON or XSD for XML, but this information is ignored when the RDF graph is constructed, as there are currently not many works that translate such schemas into SHACL. In this paper, we focus on the incorporation of XSD constraints for XML data sources in SHACL shapes. We define a translation from XSD to SHACL, and provide a corresponding system. We compare our solution with XMLSchema2ShEx which translates XSD constraints to ShEx and validate our solution against two use cases. Our solution provides the desired SHACL shapes in a reasonable time. This allows us to automatically derive SHACL shapes for some original raw data without any manual effort.

Abstract The Relational to RDF Mapping Language (R2RML) became a W3C Recommendation a decade ago. Despite its wide adoption, its potential applicability beyond relational databases was swiftly explored. As a result, several extensions and new mapping languages were proposed to tackle the limitations that surfaced as R2RML was applied in real-world use cases. Over the years, one of these languages, the RDF Mapping Language (RML), has gathered a large community of contributors, users, and compliant tools. So far, there has been no well-defined set of features for the mapping language, nor was there a consensus-marking ontology. Consequently, it has become challenging for non-experts to fully comprehend and utilize the full range of the language’s capabilities. After three years of work, the W3C Community Group on Knowledge Graph Construction proposes a new specification for RML. This paper presents the new modular RML ontology and the accompanying SHACL shapes that complement the specification. We discuss the motivations and challenges that emerged when extending R2RML, the methodology we followed to design the new ontology while ensuring its backward compatibility with R2RML, and the novel features which increase its expressiveness. The new ontology consolidates the potential of RML, empowers practitioners to define mapping rules for constructing RDF graphs that were previously unattainable, and allows developers to implement systems in adherence with [R2]RML. Resource type: Ontology/License: CC BY 4.0 International DOI: 10.5281/zenodo.7918478 /URL: http://w3id.org/rml/portal/

RDF graphs are often generated by mapping data in other (semi-)structured data formats to RDF. Such mapped graphs have a repetitive structure defined by (i) the mapping rules and (ii) the schema of the input sources. However, this information is not exploited beyond its original scope. SHACL was recently introduced to model constraints that RDF graphs should validate. SHACL shapes and their constraints are either manually defined or derived from ontologies or RDF graphs. We investigate a method to derive the shapes and their constraints from mapping rules, allowing the generation of the RDF graph and the corresponding shapes in one step. In this paper, we present RML2SHACL: an approach to generate SHACL shapes that validate RDF graphs defined by RML mapping rules. RML2SHACL relies on our proposed set of correspondences between RML and SHACL constructs. RML2SHACL covers a large variety of RML constructs, as proven by generating shapes for the RML test cases. A comparative analysis shows that shapes generated by RML2SHACL are similar to shapes generated by ontology-based tools, with a larger focus on data value-based constraints instead of schema-based constraints. We also found that RML2SHACL has a faster execution time than data-graph based approaches for data sizes of 90MB and higher.

Within ontology engineering concepts are modeled as classes and relationships, and restrictions as axioms. Reusing ontologies requires assessing if existing ontologies are suited for an application scenario. Different scenarios not only influence concept modeling, but also the use of different restriction types, such as subclass relationships or disjointness between concepts. However, metadata about the use of such restriction types is currently unavailable, preventing accurate assessments for reuse. We created the RDF Data Cube-based dataset MontoloStats, which contains restriction use statistics for 660 LOV and 565 BioPortal ontologies. We analyze the dataset and discuss the findings and their implications for ontology reuse. The MontoloStats dataset reveals that 94% of LOV and 95% of BioPortal ontologies use RDFS-based restriction types, 49% of LOV and 52% of BioPortal ontologies use at least one OWL-based restriction type, and different literal value-related restriction types are not or barely used. Our dataset provides modeling insights, beneficial for ontology reuse to discover and compare reuse candidates, but can also be the basis of new research that investigates novel ontology engineering methodologies with respect to restrictions definition.

The process of extracting, structuring, and organizing knowledge requires processing large and originally heterogeneous data sources. Offering existing data as Linked Data increases its shareability, extensibility, and reusability. However, using Linking Data as a means to represent knowledge can be easier said than done. In this tutorial, we elaborate on how to semantically annotate data, and generate and publish Linked Data. We introduce [R2]RML languages to generate Linked Data. We also show how to easily publish Linked Data on the Web as Triple Pattern Fragments. As a result, participants, independently of their knowledge background, can model, annotate and publish Linked Data on their own.

Evaluating federated Linked Data queries requires consulting multiple sources on the Web. Before a client can execute queries, it must discover data sources, and determine which ones are relevant. Federated query execution research focuses on the actual execution, while data source discovery is often marginally discussed—even though it has a strong impact on selecting sources that contribute to the query results. Therefore, the authors introduce a discovery approach for Linked Data interfaces based on hypermedia links and controls, and apply it to federated query execution with Triple Pattern Fragments. In addition, the authors identify quantitative metrics to evaluate this discovery approach. This article describes generic evaluation measures and results for their concrete approach. With low-cost data summaries as seed, interfaces to eight large real-world datasets can discover each other within 7 minutes. Hypermedia-based client-side querying shows a promising gain of up to 50% in execution time, but demands algorithms that visit a higher number of interfaces to improve result completeness.

In order to assess the trustworthiness of information on social media, a consumer needs to understand where this information comes from, and which processes were involved in its creation. The entities, agents and activities involved in the creation of a piece of information are referred to as its provenance, which was standardized by W3C PROV. However, current social media APIs cannot always capture the full lineage of every message, leaving the consumer with incomplete or missing provenance, which is crucial for judging the trust it carries. Therefore in this paper, we propose an approach to reconstruct the provenance of messages on social media on multiple levels. To obtain a fine-grained level of provenance, we use an approach from prior work to reconstruct information cascades with high certainty, and map them to PROV using the PROV-SAID extension for social media. To obtain a coarse-grained level of provenance, we adapt our similarity-based, fuzzy provenance reconstruction approach -- previously applied on news. We illustrate the power of the combination by providing the reconstructed provenance of a limited social media dataset gathered during the 2012 Olympics, for which we were able to reconstruct a significant amount of previously unidentified connections.