1 of 26

PUBLICATIONS OFFICE OF THE EUROPEAN UNION:�CELLAR, THE CENTRAL REPOSITORY OF THE EU

LLM X LAW HACKATHON�Cambridge, UK�21 June 2025

2 of 26

SUMMARY

  1. Presentation of the Publications Office of the EU
  2. What is Cellar?
  3. How is the data structured on Cellar?
  4. Which data to retrieve in Cellar
  5. APIs / Demos
  6. Numbers

3 of 26

Publication Office: OUR MISSION�AND VISION

  • make known to the public what the EU does and means and help to unlock the power of data
  • ensure that information is available as accessible and reusable data
  • facilitate transparency, economic activity and the diffusion of knowledge in all EU languages

A well-informed EU, empowered by timely�and effective access�to trustworthy information and knowledge and benefiting from all the opportunities�this brings to society�and the economy

4 of 26

Cellar

5 of 26

Dissemination: How can you use Cellar’s data?

CELL

CELLAR

Knowledge Graph

Content files

Technical data

Virtuoso/SPARQL to retrieve metadata

https://publications.europa.eu/webapi/rdf/sparql

REST API to retrieve metadata and content

RSS feeds about ingestion

http://publications.europa.eu/webapi/notification/ingestion ?{parameters}

Interfaces

6 of 26

Cellar: OP’s central data repository

6

Cellar

OP’s central data repository

Digital content

Metadata

& semantics

Multilingual content

Machine-readable

Open to all citizens for retrieval and (re)use

7 of 26

Why Cellar exists?

7

Store and disseminate (provide access to) content & metadata

Provide rich and exploitable links between the publications

Disseminate the metadata as Linked Open Data

Feed automatically the long-term archive

8 of 26

What do we publish?

8

Content

Title, abstract, dates, subject, identifier, legal information etc.

...in different file

formats &

languages

Metadata

9 of 26

Data layers in Cellar

9

Data Layer

METADATA: Structured data about the digital content

DATA: Digital Content

(documents in specific file formats and EU languages)

Data Definition Layer

COMMON VOCABULARY

AUTHORITY TABLES

THESAURI

Semantics

Structure and Formats

Knowledge Base

Knowledge

Graph

10 of 26

Linked Data and FAIR principles in CELLAR

10

**https://www.go-fair.org/fair-principles/

  1. Use URIs as names for things
  2. Use HTTP URIs so that people can look up those names
  3. When someone looks up a URI, provide useful information, using the standards (RDF/S, SPARQL)
  4. Include links to other URIs so that they can discover more things

Design Principles of Linked Open Data*

*https://www.w3.org/DesignIssues/LinkedData.html

Findable

Accessible

Interoperable

Reusable

**

11 of 26

Data structure

11

Work

Expression

Manifestation

Item

manifestation_

manifests_expression

item_

belongs_to_

manifestation

expression_

belongs_to_work

Part 3: Data

Physical/digital file level – file itself

High-level concept, an intellectual or artistic creation, for example a specific Official Journal

Language level; linguistic version of the work

Format level; a format of specific linguistic version of the work

owl:sameAs

12 of 26

What can I retrieve in Cellar ?

12

DIGITAL �CONTENT IN A �SPECIFIC LANGUAGE �& SPECIFIC FORMAT

RDF DESCRIPTION �OF A PUBLICATION

XML DESCRIPTION �OF A PUBLICATION

DESCRIPTION OF DIFFERENT PUBLICATIONS

13 of 26

How to retrieve content streams (digital files), for example a publication in PDF ?���

13

    • Retrieve digital content from its cellar id
    • Retrieve digital content from its production id
    • Retrieve digital content from the cellar id/production id, a language and a format

eng

fra

application/xhtml+xml

application/pdf;type=pdfa1a

Accept:

Accept-language:

14 of 26

How to retrieve metadata of a publication ?���

14

  • Retrieve metadata for an object (work, expression, etc) in RDF format.
  • Retrieve metadata for an object (work, expression, etc) in XML format
  • Retrieve all the identifiers of a specific resource (sameAs)

application/rdf+xml;notice=tree

application/rdf+xml;notice=non-inferred-tree

application/rdf+xml

application/rdf+xml;notice=non-inferred

application/xml;notice=tree

rdf notice:

Accept:

xml notice:

Accept:

15 of 26

Accessing metadata using SPARQL?���

15

prefix cdm: <http://publications.europa.eu/ontology/cdm#>

SELECT distinct ?act ?sig

where {

?c_act owl:sameAs ?act.

?c_act cdm:official-journal-act_date_publication "2023-10-02"^^<http://www.w3.org/2001/XMLSchema#date>

FILTER (regex(?act,'/oj/')).

?c_exp cdm:expression_belongs_to_work ?c_act.

?c_manif cdm:manifestation_manifests_expression?c_exp.

?c_sig cdm:signature_digital_signs_manifestation?c_manif.

?c_sig owl:sameAs?sig.

}

16 of 26

Should have to use API or VIRTUOSO to retrieve metadata?

API

  • I want to extract the full direct metadata of a publication (I can also focus on a language or format)
  • I want to extract the full inferred metadata of a publication

� I need to know the URL (cellar id or psi id) of the publication

VIRTUOSO

  • I want to list publications according to some criterias (collection, topic, format, etc.)
  • I want to browse relations between publications
  • I want to extract direct metadata of a publication

16

17 of 26

Magnitude of data in Cellar

17

Digital content size

    • 48 million files, and growing

Virtuoso size

    • >1.6 billion piece of data (triples)

~29 million hits per day in 2025

18 of 26

THANK YOU

emmanuelle.GAILLARD@publications.europa.eu

abdelfettah.FELIACHI@publications.europa.eu

19 of 26

Annex

19

20 of 26

21 of 26

  • online access to EU and national law
  • easily findable, accessible and transparent
  • EUR-Lex, EU Law in Force, OP Portal and N-Lex
  • over 2,000 easy-to-read summaries

22 of 26

EUROPEAN DATAdata.europa.eu

  • open data news, studies, trends, stories, and academy
  • data from EU institutions and public sector
  • cca 1.9+ million datasets

EU TENDERS�ted.europa.eu

  • transparency & access to business opportunities in the EU and beyond
  • over 3,000 notices per day
  • annual total value:�cca 815 billion EUR
  • primary source of results from the projects funded by the EU's programmes for research and innovation
  • foster open science, create innovative products and services and stimulate growth
  • oldest EU website founded in 1994

EU RESEARCH RESULTS�cordis.europa.eu

23 of 26

EU PUBLICATIONS�op.europa.eu/publications

  • reports, studies, information brochures, magazines and other publications from EU institutions, agencies and bodies
  • ‘In your language’: publications in 24 official EU languages
  • publications organised in thematic bundles

24 of 26

EU WHOISWHO�whoiswho.europa.eu

  • The official directory of the EU
  • Over 17,000 persons registered
  • Contact details of all EU institutions, agencies and bodies (weekly updates)

25 of 26

  • Ontologies, authority tables, thesaurus used in EU publications
  • Easy and standardized way to publish
  • Facilitates search and access to EU content
  • Supports harmonization across languages and borders
  • Powered by semantic technologies and interoperability principles
  • Easy and standardized way to publish

26 of 26

Three WEMI hierarchies in Cellar

26

Work-expression-manifestation-item (WEMI) hierarchy for digital content (content streams) representation

Dossier-event hierarchy for dossiers and sequence of events related to dossiers (from initiation to conclusion)

Agent hierarchy for representing countries, persons, organisations, groups, committee, institutions, courts etc.