The participation of the SEMIC Team at the Endorse Conference

Image of the ENDORSE conference 2025 by the Publications Office of the European Union

With the Artificial Intelligence Act, entering into force on the 1st of August, the 3rd edition of the ENDORSE conference fell at a timely moment to discuss the opportunities and challenges of Artificial Intelligence (AI) for reference data and semantic technologies.  

ENDORSE, the European Data Conference on Reference Data and Semantics, took place on 8 and 9 October 2025. The goal of the ENDORSE conferences is to enable sharing of knowledge, experiences, and best practices of interoperable reference data and semantic technologies. 

This year, ENDORSE’s theme was “Reference Data and AI: Transforming Data into Action across Borders and Languages.” The Semantic Interoperability Community (SEMIC) was also present, with three main presentations by our team members, and various networking opportunities. Among the other presentations, our speakers covered topics from digital-ready policymaking, semantic model discovery and harmonisation, new AI tools to enhance interoperability, to MLDCAT-AP and interoperability in machine learning models. In the coming sections, you will get information on the SEMIC-related presentations at ENDORSE this year. 

Claudio Baldassarre presenting at the ENDORSE conference 2025, picture by Interoperable Europe 

The SEMIC team introduced the Semantic Registry, a central portal designed to advance cross-border semantic interoperability by making domain data models and their semantics discoverable, assessable, and reusable across Europe. 

The Semantic Registry’s goal is to enable consistent selection and reuse of data models, promoting interoperability across sectors. The Registry empowers civil servants to make informed choices and supports the convergence of design and adoption across Member States and EU agencies. 

Claudio highlighted a core challenge: semantic models are dispersed across multiple repositories, making them difficult to find and assess. The Semantic Registry solves this by bringing asset metadata into a trusted catalogue helping users discover and adopt models. Users can also compare models based on adoption, maturity, technical details, and connection with other standards. This aims to directly address semantic interoperability during the design phase.   

The next phase is a pilot to operationalise the Registry and develop user interfaces and APIs. Claudio invited experts involved in modelling activities for the public sector and interested individuals to join the pilot, contribute feedback, and help shape a European data ecosystem where semantic assets are easier to find, reuse, and trust. 

To organise a pilot or have more information, you can contact the SEMIC team. More information can be found in their ENDORSE presentation. 

Emilien Caudron (left) and Georges Lobo (right) presenting at the ENDORSE conference 2025, image by Interoperable Europe

Achieving semantic interoperability, ensuring data is exchanged with shared unambiguous meaning, is critical but is complex and demands deliberate effort from the data modellers. To address this, Emilien Caudron and Georges Lobo demonstrated on stage the SEMIC AI Assistant, an innovative tool that leverages Large Language Models (LLMs) and a Retrieval Augmented Generation (RAG) pipeline to help users find and reuse the right concepts, aligned with recognised standards and reducing duplication in data modelling. Future development will include predicting semantic tags, aligning models, and generating conceptual semantic models.  

Emilien and Georges showcased on stage a practical use case of the assistant, uploading a data model, receiving visual overviews, and quickly receiving feedback on gaps or overlaps in the model.  

Interested in testing the SEMIC AI Assistant? Contact the SEMIC team for an opportunity to participate in pilot phases. More information can be found in their ENDORSE presentation. 

Jiste De Cock (left), Emidio Stani (centre), and Ine Weyts (right) at the ENDORSE conference 2025, image by the Publications Office of the European Union

This presentation introduced MLDCAT‑AP (Machine Learning Data Catalogue Application Profile), a shared “catalogue” standard for describing machine learning models and datasets. As machine learning becomes increasingly integral to public services, there is a growing need for a shared catalogue such as MLDCAT-AP. MLDCAT-AP build on the widely used DCAT-AP metadata standard, adapting it for the specific needs of AI and machine learning.  

Initially developed by the SEMIC team with OpenML under Interoperable Europe, version 3.0.0, which was released on the 1st of October 2025, aimed to broaden its scope and strengthen alignment with policy. This latest version of MLDCAT-AP incorporates elements of the AI Code of Practice, supporting modelers with compliance and risk mitigation, and updated alignment to DCAT-AP 3.0.1. 

More information can be found in their ENDORSE presentation. 

Interested in conducting a Pilot with the SEMIC team? Contact us! Interested in improving MLDCAT-AP? Follow this MLDCAT-AP GitHub link to share your feedback with the SEMIC team. 

 Davide Audrito (left) and Alessio Nardi (right) at the ENDORSE conference 2025, image by Interoperable Europe 

This talk presented the Legislative Financial and Digital Statement (LFDS) as the Commission’s lead approach to digital‑ready policymaking and explained how a structured model makes information that is now mostly free‑text easier to find, reuse and analyse. It showed how the template was designed, how it connects with existing legal standards, and how early testing, using a mix of manual and semi‑automated methods, lightened the workload for policy officers. 

More information can be found in their ENDORSE presentation. 

Join the Interoperable Europe Community and the mailing list to get information of future updates.