Programme

The EU Vocabularies website, managed by the Publications Office of the European Union, provides essential semantic resources and tools to enhance data management and interoperability across EU institutions and emerging European Data Spaces. These resources, including authority tables, code lists, thesauri, taxonomies, and ontologies, ensure consistent and standardised data exchange while fostering a unified digital environment. Key tools include VocBench and ShowVoc, which facilitate semantic data management. VocBench is a web-based application for creating, enriching, and publishing multilingual controlled vocabularies, supporting collaborative workflows and adhering to semantic web standards like SKOS, RDF, OWL, and SPARQL. ShowVoc complements VocBench by providing an intuitive interface for browsing and aligning data, making controlled vocabularies accessible to non-technical users. Both tools ensure data consistency, usability, and interoperability. The Cellar repository and the EU Vocabularies platform serve as official publication channels, offering APIs and SPARQL endpoints for machine-readable data. Authority tables standardise terminology across 200+ topics, supporting seamless categorisation and cross-domain integration, vital for European Data Spaces in areas like health, mobility, and procurement.
Adhering to the FAIR principles (Findable, Accessible, Interoperable, Reusable), EU Vocabularies ensures data accessibility and reuse. Business Collections provide domain-specific semantic frameworks, fostering standardisation and collaboration in business processes. By providing structured, interoperable data, EU Vocabularies enables effective cross-sectoral and cross-border collaboration, playing a pivotal role in advancing European Data Spaces and supporting innovation across the EU.

The Common Assessment Method for Standards and Specifications (CAMSS) was developed by the European Commission having the interoperability between Digital Public Services (DPS) in mind. Software solutions that use standards, and largely adopted specifications are good candidates to be chosen as components of Interoperability Environments (IE). The identification and selection of a specification – be it a de facto or de jure standard or not – to support the implementation of a solution is not a straightforward decision, though; notably when the interoperability of the solution within its IE and with other IEs is at stake. Hence, in the context of European DPS, assessing these specifications becomes a must, and recently has become a legal requirement (cfr. Interoperable Europe Act, Article 3). CAMSS provides “scenarios” to assess the interoperability and compatibility of such specifications.

CAMSS (CAMSS-ELIS) and ENDORSE can benefit mutually in different ways. ENDORSE can take leverage of  already existing assessments and documented links to EIRA ABBs. ELIS could use the knowledge and expertise of the ENDORSE community to identify new types of specifications, notably controlled vocabularies and reference data that are recommended or extensively used to enable interoperability between European public administrations. Ultimately, tools used by ENDORSE’s users may integrate or interoperate with the suit of tools and solutions offered by the Interoperability Architecture initiative mentioned above. Additionally, ENDORSE could recur to the CAMSS as a Service team to perform interoperability and incompatibility assessments of reference data and other types of specifications identified by the ENDORSE Community.

The Publications Office ‘Metadata and Identifiers Sector’ (A.2.002) creates metadata to describe general publications in the domain of publications. To provide multilingual descriptions, translation tables (ATTO tables, now TLT) are used since 2003. These tables are managed internally and, where possible, their values are mapped to corresponding ones in Authority tables (AT) available through Cellar. Despite the usefulness of the ATTO tables, maintenance and linked data compatibility have been two major issues.

Maintained by the cataloguing team, ATTO tables contain codes and many translations. Codes are used during cataloguing, while labels are used to support multilingual displays. On the other hand, ATs are reference data with clear semantics, definitions, labels in different languages, mappings to values existing in local ATTO tables, and permanent URIs for identification and reference.

An analysis by a designated team of A.2.002 librarians recommended phasing out ATTO tables in favour of ATs. The examination of ATTO tables and their individual values began in 2021 and concluded by mid-2024. In addition to cataloguers, production teams were consulted to identify new concepts. Close collaboration with the ‘Reference Data and Style Guide Sector’ (A1.002) enabled the determination of semantics and definitions. This effort resulted in the development of six new ATs and the update of nine existing ones.

This presentation will detail the analysis process, the challenges encountered, and the lessons learned. Full integration of the ATs into everyday cataloguing business is anticipated by the end of 2025, along with the deployment of the new ‘linked data-enabled’ cataloguing tool.

Welcome to a few minutes with practical use of our (inter)national standards for metadata descriptions!

We'll look into events, public services and concepts from a real-life use case. And how we can describe these with machine-readable knowledge, as steps in establishing F.A.I.R. data for interoperability.

The results of this talk will be interoperable metadata, ready to go to data.europa.eu, and with that ready for
usage by you and others! For without interoperable metadata, you won't have interoperable data.

The European Union (EU) legislation includes Reporting Requests (RR) on a diverse range of stakeholders, including public institutions, requiring the regular submission of information to other entities to allow such addresses to perform their tasks. The European Parliament in the “Review clauses in EU legislation adopted during the first half of the ninth parliamentary term (2019-2024)” , states that the 60% of the EU legislation includes RR, with a more frequent timeframe of 1 year, repeated for 4 years. For this reason, the European Commission has explored experimental approaches to mitigate the complexity of this tasks using legal ontologies and AI tools . The Study on Regulatory Reporting Standards (SORTIS) project  aims to
create the RRMV ontology for the EU legislation. We intend to present RRMV and the evolutions.

This presentation explores the integration of AI and semantic technologies in legislative drafting, focusing on the development of an "augmented LEOS" system. Based on the results of a comprehensive European Commission study, it highlights smart functionalities that enhance drafting efficiency and accuracy while ensuring compliance with legal standards. Central themes include the use of semantic specifications, multilingual intelligence and reference data to create machine-readable legal texts. Key contributions include a detailed categorization and prioritization of 61 smart functionalities, emphasizing
their role in verifying legal consistency, linguistic support and automated drafting. Among others, the study integrates hybrid AI approaches and Large Language Models to address complex legal drafting tasks. The proposed high-level roadmap outlines steps from technology exploration to deployment.By aligning emerging AI technologies with open standards and datasets, this work paves the way for interoperable digital governance solutions in the field of advanced law making. The presentation will conclude with lessons learned from piloting AI-enhanced tools in legislative environments, offering insights into creating scalable, transparent and user-centric systems for the next generation of digital policymaking.

As per January 2025, the main instrument of digital-ready policymaking in the European Commission is the Legislative (Financial and) Digital Statement, which compiles essential information on the digital aspects of a policy initiative. 
To be able to tap into the richness of such information initially provided in plain text, and to enable further use cases, adopting a structured information approach is key. The paper focuses on presenting how the model was developed, the main design rationales, the use cases that the model supports, as well as the relations with other existing semantic specifications in the legal domain.

Additionally, the article presents intermediate results obtained while building such reference data, and the manual and semi-automated approaches taken to reduce the burden on the policy officer responsible to complete it.

The specific technical details needed for the smooth implementation and enforcement of the AI Act require
specification of additional implementing acts, standards, codes of practice, codes of conduct, guidelines, templates and benchmarks. These will require ongoing consensus building either in external bodies, such as the European Standards Organisations which was requested to develop harmonised standards for the Act, or fora convened by the Commission to develop codes of practice, codes of conduct, guidelines, templates, benchmarks and implementing and delegated acts. The European Commission, the AI Office and the networks of national competent authorities, rights protection bodies, certification bodies and market surveillance bodies will face a major challenge in building consensus efficiently, adapting it to accelerating AI advances and offering a consistent and innovation-friendly enforcement.

To address this challenge, the implementation and enforcement of the AI Act must rapidly and comprehensively adapt and adopt existing mechanisms for sharing public interest information as open data adopting Findable, Accessible, Interoperable and Reusable (FAIR) information principles. It presents proposals and gaps in open data vocabularies, taxonomies and ontologies based on RDF should be considered to support key reference data for implementing the AI Act.

This will present and share the experience gathered during the Ref2Link project. Ref2Link is a multilingual EU legal or administrative references extraction technology developed by the European Commission Legal Service and supported by the Digital Europe Program. Pushing it to its limits, given the complexity of the multilingual scope, Ref2Link uses well established rule-based text pattern recognition to extract references, adhering to standard natural language citation or identifier formats - such as the European Legislation Identifier (ELI) or the European Case Law Identifier (ECLI). Implemented use cases cover a horizontal variety of business needs such as automation of links or footnotes/endnotes, text categorisation, data capture and quality control. Several integrations in web or desktop IT systems deliver a direct benefit to business users.

The session will highlight how, by using references as gateways to semantic web repositories such as the EU Publication Office Cellar, Ref2Link not only improves the accuracy of detection beyond text pattern detection, but also greatly enhances user experience by pulling linked data or allowing to browse through legal knowledge graphs. Current AI based enhancements under development will also be presented.

Automated extraction of structured information from Eu-Acquis, the large collection of common rights and obligations that constitute the body of EU law, is paramount for enabling policy and decision-makers across the EU to access sought-after precise information in a time-efficient manner. EU-Acquis is stored in the common data repository of the Publications Office of the European Union, i.e., Cellar. We introduce the generic task of Information Provision Activity Requirement Extraction (IPARE) which focuses on identifying in the text of EU-acquis text fragments which introduce an obligation or a possibility of information provision activities and extracting structured key information therefrom. As the EU acquis evolves over time, IPARE requires continuous monitoring effort. IPARE consists of two subtasks, i.e., detection and extraction. While the former employs binary classification to identify if an information provision activity requirement is mentioned in a given text fragment, the latter aims to extract primary information on providing entity, target entity, frequency, and deadline of the provision activity.

We report on our experiments and present a comparative evaluation of various Natural Language Processing technologies for the tasks at hand. More specifically, we create a novel benchmark corpus of text snippets with information provision activity requirements sampled from EU-Acquis, and compare knowledge-based, classical machine learning, and LLM-based approaches for the task at hand, with the ultimate goal of answering the question of which technology might be the most suitable and elaborate on the eco-friendliness thereof.

We will show how the SHACL Play suite of tools (https://shacl-play.sparna.fr) can be leveraged to setup a
model-driven approach in knowledge graph projects.

In particular SHACL Play enables to use SHACL specifications as the source of truth to automate certain
tasks in a data flow : data validation, end-user documentation generation, dataset statistical analysis,
JSON-LD context and JSON schema generation, Sparnatural query UI configuration, etc. Facilities are also provided to acquire SHACL specification from the analysis of an RDF dataset, or to edit SHACL specifications easily in Excel spreadsheet templates.

We will illustrate how it has been applied for the European Parliament Open Data Portal. In particular if was applied at three steps in the data flow : at data acquisition step, to validate the structure of individual data source migration; at the central knowledge graph, to specify the structure of the entire knowledge graph; at dataset dissemination step, to specify the structure of generated datasets.

The semiconductor industry’s complexity necessitates precise communication and collaboration among diverse stakeholders. Misunderstandings in terminology can lead to inefficiencies and disruptions. The Ontology Curation Portal (OCP) is a comprehensive platform designed to address these challenges by facilitating collaborative ontology development, editing, and curation, specifically for the semiconductor domain. OCP provides tools for semantic precision, offering three ontology visualizations tailored to the needs of knowledge engineers, domain experts, and business users. This enables the effective communication of complex relationships. The platform integrates the Vocabulary Development Service (VDS), which standardizes terminology, resolves ambiguities, and supports machine-readable formats (RDF) for interoperability. Additionally, the OCP supports multi-stakeholder collaboration, ensuring inclusivity and scalability in semantic specifications. This presentation will demonstrate how the OCP empowers organizations to create domain-specific semantic models that enhance interoperability, improve machine-to-machine communication, and drive innovation in the semiconductor sector. A full extended abstract will be submitted for inclusion in the conference proceedings.

Data in the telecommunications industry is generally complex and hard to categorise, due to crossing the boundaries between many domains and themes including organisational, geospatial, technical, socio-economic and commercial. As such, there are significant challenges in data integration and knowledge representation.

The authors, both working in a telecommunications analyst and data company, present the Common Telecoms Ontology (CTO), a comprehensive semantic framework designed to structure and standardize telecommunications (particularly broadband) data. The CTO provides a unified model for representing relationships between organizations (ISPs and network operators), physical infrastructure (networks and
link technologies), commercial information (tariff data) and geospatial information (locations), while maintaining temporal validity and geographical context.

Our implementation demonstrates how semantic technologies can address real-world telecommunications data challenges, including tracking network ownership, service provision
relationships, and infrastructure deployment. To the best knowledge of the authors, this is the first attempt at an ontology for telecommunications data which is not overwhelmingly limited to the technical domain.

To allow compatibility, the CTO ontology incorporates entities from established namespaces such as FOAF (Friend of a Friend) and DC Terms (Dublin Core) and is compatible with the DCAT ontology for European data, while introducing its own telecommunications-specific classes and properties. https://github.com/Point-Topic/cto-ontology

Facilitating the reuse of standardised concepts, definitions, or properties is a significant challenge for achieving semantic interoperability. To address this challenge, the use of generative artificial intelligence and Retrieval Augmented Generation has been explored. By linking a generative AI model to a knowledge base of interoperability standards (CPSV-AP, CPOV, CCCEV, Open Data directive, ...) and data models (Linked Open Vocabularies, schema.org), an AI assistant was developed. This assistant enables users to upload, interact with, and visualise data models, and more crucially, allows for the modification of models through the reuse of established concepts and definitions. The approach aims to enhance the efficiency and accuracy of data modelling, reduce redundancy, and improve data quality, thereby promoting collaborative and compliant data management solutions. Although currently at the proof-of-concept stage, this tool illustrates the potential of AI in advancing data modelling and interoperability, with future developments focused on expanding its capabilities and exploring additional use cases.

The Mapping Workbench (MWB) offers a transformative approach to managing the growing complexity of semantic data mapping in public-sector projects. With evolving data schemas, semantic consistency requirements, and multi-stakeholder collaboration challenges, MWB provides an integrated platform for conceptual and technical mapping, validation, and automation.

This presentation explores the motivations behind MWB’s development, rooted in the need for streamlined workflows, improved productivity, and semantic precision. MWB supports the entire mapping lifecycle, from aligning business and technical stakeholders to generating validated, machine-executable mappings. Key features include SHACL and SPARQL validation tools, an IDE-like interface for automation, and collaborative workflows that ensure consistency across evolving datasets.

The session will briefly highlight MWB’s successful application in mapping eForms notices to the eProcurement Ontology (ePO), a large-scale public procurement initiative. By addressing real-world challenges, MWB has proven its value in creating high-quality, scalable semantic data mappings.

Attendees will gain insights into how MWB addresses the demands of interoperability, enhances collaboration, and ensures semantic quality, making it a powerful tool for advancing semantic technologies across various domains.

As Artificial Intelligence (AI) and Machine Learning (ML) enter increasingly more and more diverse domains, the need for standardised metadata has risen to ensure interoperability, transparency and compliance. The lack of a standard practice for describing ML models and datasets poses a significant challenge. Without a consistent approach, discovering, understanding, and reusing existing machine learning models and datasets becomes difficult, resulting in duplicated efforts and inconsistencies. Recognising these challenges, SEMIC, in collaboration with OpenML, aimed to develop a specification that would effectively address these issues. The objective was to create an application profile for describing machine learning models, datasets and other relevant concepts. This resulted in the creation of the Machine Learning Data Catalogue Application Profile (MLDCAT-AP).

As part of the Interoperable Europe initiative, MLDCAT-AP addresses challenges in sharing and reusing ML models and datasets across diverse domains and organizations. By leveraging standardized data practices, MLDCAT-AP facilitates the integration of machine learning models into broader data ecosystems, promoting transparency and reproducibility in ML research and development. MLDCAT-AP represents a significant step forward in enhancing the interoperability and discoverability of machine learning models and encourages compliance with the AI Act.

The European strategy for data aims to create a single European data space acting a genuine single market for data, open to data from across the world along with its forerunner, the revised Open Data Directive i.e. High Value Datasets which aims to boost sharing of data falling into categories of open public sector data. In addition to this, sharing of sensitive data possessed by public institutions should be supported in a trusted manner. The aforementioned regulations create new challenges to data.europa.eu which aim is to give access and foster the reuse of European open data among citizens, business and organizations. The portal collects the metadata from the data providers even when it contains deviations from the common data model (DCAT-AP).

With the scope of improving the quality of the metadata and data, the portal constantly assesses the quality of the metadata errors and communicates them back to the owners of the datasets. Hosting datasets based on extensions of the common data model in the context of data spaces and HVDs raises for data.europa.eu a number of challenged in a semantic approach.

Some of those include among others: 

• once only-effort for publishers of datasets,
• semantic integration (a shared and formalized understanding on the meaning of the used terminology/concepts) & effective mapping of provided metadata
• release of more and better-quality metadata and data ensuring thus, effective findability, accessibility and reusability from citizens,organisations & companies.
• Possible adaptation of aggregation techniques
• Use of artificial intelligence tools that might assist towards this direction.

The ISO & GeoDCAT-AP Metadata Implementation Pilot is a collaborative action supporting the encoding of geospatial metadata under the INSPIRE Directive using the GeoDCAT-AP extension of the W3C RDF vocabulary for data catalogues (DCAT). As part of the GreenDataForAll initiative, it aims to modernise and simplify European environmental data sharing (INSPIRE) towards the nascent European Green Deal Data Space, besides the alignment to the High-Value Datasets Implementing Regulation.

The pilot evaluates the suitability and quality of the GeoDCAT-AP 3.0.0 specification and its accompanying XSLT transformation, testing the conversion of geospatial metadata from ISO 19139/INSPIRE into GeoDCAT-AP. Experts from nine European countries (Belgium, the Czech Republic, Denmark, Finland, France, Italy, the Netherlands, Spain, and Slovakia) participate, alongside European Commission services (SEMIC group of DIGIT, CNECT, ENV, Publication Office, and Joint Research Centre).

The action facilitates the integration of INSPIRE reporting procedures into the Open Data High-Value Datasets reporting flow towards the European Data portal, thus reducing implementation efforts under both legal initiatives. Expected benefits include improved quality-tested metadata and tools, gaining hands-on experience on the new specification and its implementation, while minimising the burden on Member States.

This presentation will provide a comprehensive overview of the pilot from business and technical perspectives, highlighting its main results, achievements, and lessons learnt applicable to other data spaces where legacy standards are in place. By sharing these insights, the pilot's findings are expected to have significant impact on public sector European data sharing in the coming years.

The World Wide Web Consortium (W3C) was founded 35 years ago to standardize web technologies and lead the Web to its full potential. As our personal and professional lives have become more and more digitized, the question of defining and proving one’s identity on the Web has become a critical topic, and has been addressed in various W3C recommendations.

In this presentation, we will review the state of standardization activities in W3C related to identity and credentials. We will present the current and upcoming recommendations, as well as areas where they have been successfully deployed. The presentation will also focus on the link between these recommendations (in particular W3C Verifiable Credentials) and semantics technologies, and show the benefit they can have on interoperability across systems, communities and borders.

The publications Office of the European Union is mandated to build and implement the EU publication process. This presentation will describe how the introduction of a new format such as AKN4EU, with the interoperable aspect as a main goal, does impact the publication process. Built on the IMFC Common Vocabulary (https://op.europa.eu/en/web/eu-vocabularies/cov), Akoma Ntoso for European Union (AKN4EU) is a machine-readable structured format for the exchange of legal documents in the EU decision-making process. It is based on XML, and more specifically on Akoma Ntoso (https://en.wikipedia.org/wiki/Akoma_Ntoso), an OASIS standard (https://www.oasis-open.org/standards), and serves as the specification for all future exchanges of legal documents.

The scope of the current version includes, primarily, legal acts related to the ordinary legislative procedure (regulations, directives and decisions), as well as the respective legislative proposals. The integration of this new structured format has an impact on the organisation of dataflow interaction across EU institutions. This presentation will focus on the new interinstitutional framework description following the integration of this new format. It will also cover a set of existing tools which are required to support the milestone change for edition, validation and conversion of EU legal documents.

The digital transformation of legislative documents presents significant opportunities in improving the accessibility, efficiency, and interoperability of legal information across the European Union. An excellent example of such transformation is AKN4EU, an innovative XML-based, machine-readable format facilitating the seamless exchange and processing of legislative documents. The presentation will commence with an introduction to AKN4EU, highlighting its transformative potential. Central to our discussion will be the concept of interoperability, AKN4EU as a key enabler of interoperability in EU legislation, the interinstitutional process of developing the AKN4EU standard, and the different roles of the Publications Office as a neutral facilitator, independent stakeholder as well as guardian of excellence in that process. The presentation will conclude with an exploration of the benefits that AKN4EU brings to the table, especially when combined with other state-of-the-art technologies such as artificial intelligence. These benefits include increased accessibility, the facilitation of collaborative legal drafting, and the ability to reuse legal documents with ease.

The Semantic Web has made it possible to publish and link datasets using common standards, promoting interoperability and reuse, and enabling federated access to information. Our system, VocBench, has supported collaborative yet controlled development and maintenance of ontologies, thesauri, lexicons, and Semantic Web datasets in general. However, it has not addressed the pressing need for handling tasks related to the automatic processing of unstructured content, particularly text documents. We present SELEN (Semantic ELicitation ENvironment), which fills this gap as a platform for orchestrating content analysis for RDF data elicitation. As SELEN approaches its first release, we outline its scope and discuss its design and intended benefits.

After presenting SELEN's requirements, we illustrate how we met them by designing an extensible architecture and an associated REST API described in an OpenAPI documentation. Rather than being a monolithic solution, SELEN is open to different implementations that can be plugged into extension points, while supporting training on different semantic datasets and document bases. In this way, SELEN supports different use cases as well as different, potentially interchangeable algorithms within a predefined framework, including state-of-the-art AI models such as Large Language Models (LLMs). SELEN is a natural companion to VocBench, which has been enhanced with a user interface that allows humans to be brought into the loop to validate and refine the automatically generated suggestions. We evaluate the platform against relevant non-functional requirements for availability, usability, and maintainability. Finally, we discuss the benefits of SELEN for maintaining reference data and for applications that use it.

The Open Standards for Linked Organisations (OSLO) is an initiative by the Flemish Government designed to foster the data economy by providing open semantic data standards. Over its more than 10 years of existence, OSLO has gained the trust of the Flemish public sector as the go-to approach for creating semantic data bridges that span organisational borders.

In this presentation we will highlight a number of challenges this growth has brought and which improvements are made to address these challenges.

Cultural heritage collections are invaluable for preserving history, but their metadata can perpetuate harmful biases and stereotypes, creating ethical and accessibility challenges. The DE-BIAS project addresses these issues by developing an AI-based tool to detect and curate harmful language in cultural heritage metadata. A key outcome of the project is the DE-BIAS vocabulary, a structured resource of derogatory and contentious terms created and curated within the project. This vocabulary, now published in the EU Vocabularies, provides a standardized and accessible foundation for identifying problematic language across diverse cultural contexts, in five European languages: English, German, Italian, French, and Dutch.

The DE-BIAS tool leverages the vocabulary alongside advanced Natural Language Processing (NLP) techniques, including rule-based and machine learning methods, as well as Large Language Models (LLMs), to flag harmful terms and provide culturally sensitive context and suggestions. The tool has been integrated with platforms like Europeana and MINT, and is also available as a standalone service, ensuring widespread applicability in cultural heritage workflows.

This presentation details the methodology behind DE-BIAS, focusing on the creation and curation of the vocabulary, the development of the AI tool, and the ethical challenges of balancing historical preservation with linguistic sensitivity. By fostering inclusivity and reducing bias, the DE-BIAS project sets a benchmark for ethical metadata practices, enabling cultural heritage institutions to ensure their collections remain accurate, respectful, and accessible to diverse audiences worldwide.

In 2015 the Dutch Digital Heritage Network (NDE) was created to improve the access to cultural heritage information. Supported by the Ministry of Education, Culture and Science, NDE runs a long term program aimed at increasing the social value of information held by cultural institutions by making their collections more visible, usable and sustainable. The NDE-program works on creating a large-scale, decentralized network infrastructure for exchange and reuse of digital heritage information. With this work the foundations for a Dutch data space for cultural heritage information are being created. This national data space is the pivoting infrastructure for exchanging information with the European common data space for cultural heritage information (as deployed by Europeana) and to national and international research infrastructures for social sciences and humanities. Working together with commercial vendors NDE has been able to create an ecosystem that provides solutions for collection management which support Linked Data publication functionality. A DCAT compliant dataset registry was created that crawls the network and collects the published dataset descriptions. And a federated search engine was developed to perform a real time search in terminology sources to find the available identifiers for standard definitions such as places, persons, subjects. Together with service providers and platform developers the NDE-programme works on generic ways to create end user services based on standard APIs and functionality. This presentation shows the results of the concrete work that is being done in the Netherlands for building the Dutch data space for digital heritage.

There are ten Belgian Federal Scientific Institutions (FSIs) that curate and preserve numerous scientific, artistic and historical heritage. Linked to collections items are so-called authority data such as its  creator as well as related metadata such as birth date or birth place. This is important reference data, among others for correct attribution or the identification of works that leave copyright and enter public domain (based on known death dates). Several FSIs focus on cultural heritage and hence there is an overlap of said person authority data in the different FSIs. Yet, this data is currently curated by each FSI separately, meaning that different standards are used and that semantics differ. This not only results in duplicate curation work, but also in non-connected data silos that are difficult to access by researchers and the public.

In this presentation we introduce the MetaBelgica project, coordinated by the Royal Library of Belgium (KBR) with the Royal Museums of Fine Arts, the Royal Institute for Cultural Heritage and the Royal Museums of Art and History. This project aims to unlock more than one million authority data records with the Open Source software Wikibase. We will explain why we have chosen for a central database over a decentralized solution. In addition, we will report on our data integration strategies that require manual curation due to incomplete or uncertain metadata. This platform makes reference data available under a persistent identifier and will serve as a single trustworthy source of Belgian Entity data related to Cultural Heritage.

The DataGEMS project aims to improve the accessibility and interoperability of digital resources by creating tools to generate detailed metadata across different data, enhancing dataset integration and discoverability. Using historical German encyclopedias from the EncycNet project as a case study, we show how semantic enrichment via Wikidata can facilitate insights into this kind of historical language data. Six general (“conversational”) encyclopedias from the 19th century were linked to Wikidata using textual similarity, leveraging its taxonomy for a quantitative topic analysis. Two different analyses on broader vs. more fine-grained topics revealed

1. a shift from abstract concepts to factual knowledge, with increased focus on people and places driven by globalization and industrialization, and
2.  a decline in religious content relative to scientific topics, reflecting a broader societal shift towards empirical knowledge.

With DataGEMS, we want to further facilitate the exploration of semantic shifts among others, such as the evolution of terms and changes in knowledge distribution. It aims to support advanced research into historical networks of ideas, propagation of concepts, and the structuring of knowledge. Ultimately, encyclopedias are valuable tools for understanding historical knowledge and can independently enrich other historical datasets as referential resources.