Open data · Semantic Web
The knowledge graph behind every answer
Every MP, speech, act and vote you read on ParliamentRAG lives in a knowledge graph: a network of facts linked to the open data of the Italian Chamber, the Senate and the European Union. You can download the whole graph from this page.
The graph in numbers
Chamber of Deputies
414
deputies
46,844
floor speeches, full text
705
floor sittings
35,773
parliamentary acts
17,295
electronic ballots
6.9M
individual votes, one per MP
Senate of the Republic
212
senators
17,904
floor speeches, full text
464
floor sittings
15,444
parliamentary acts
8,338
electronic ballots
1.2M
individual votes, one per MP
Shared
1,716
EuroVoc topics linked to acts
863,834
RDF triples in the graph export
Sources: open data of the Chamber of Deputies (dati.camera.it), the Senate of the Republic (dati.senato.it) and the EU EuroVoc thesaurus. The graph is updated with every new sitting.
How the data links up
[a]
Everything has an address
An MP is not a row in a private table: it is an official dati.camera.it or dati.senato.it URI, the same one the Chamber or the Senate uses. Whoever cites that address is talking about the same person, with no ambiguity.
[b]
Shared vocabularies
To say “person”, “organisation”, “topic” we use standard vocabularies (FOAF, W3C ORG, SKOS) instead of inventing our own names. Our data joins with archives, newspapers and researchers without any conversion tables.
[c]
Elementary facts
Knowledge is broken down into minimal subject-predicate-object sentences: “Aiello, is a member of, MoVimento 5 Stelle”. A machine reads them effortlessly and combines them to answer complex questions.
A deputy, in triples
The triples below come straight from the dump: they describe deputy Davide Aiello, his record and his group membership.
The URI is the official one — the identifier is the same as dati.camera.it: our data and the Chamber's data link up by themselves.
Universal vocabularies — foaf:Person is the standard way of saying “person” on the web of data; ocd:deputato is the Chamber's own open-data vocabulary.
Membership is a dated fact — group membership is an object with a start and end date (org:Membership): the graph also remembers who switches groups, and when.
The same language as official data
Each concept in the graph is aligned to a standard vocabulary. The table is the same one in the paper's appendix, and it is what makes the data readable by any system that knows those vocabularies.
| In our graph | Standard vocabulary | What it guarantees |
|---|---|---|
| Deputy | foaf:Person + ocd:deputato | the same official Chamber record, reusable anywhere |
| Parliamentary group, committee | org:Organization (W3C) | the W3C standard for describing organisations |
| Group membership | org:Membership | a fact with a start and an end: party switches stay on record |
| Topic of an act | skos:Concept — EuroVoc | the EU's multilingual thesaurus: the same topics in 24 languages |
| Sitting, debate, speech | ≈ Akoma Ntoso | the structure follows the international standard for parliamentary acts |
| Ballot and individual vote | ocd:votazione / ocd:voto | the open-data vocabulary of the Chamber of Deputies |
| Data provenance | PROV-O | every build of the graph declares when and from which sources it was made |
Download & reuse
The whole graph is exported as RDF, the open format of the Semantic Web: readable with any standard tool (Apache Jena, rdflib, GraphDB…) and queryable in SPARQL.
Data updated as of August 4, 2026
The knowledge graph
Turtle · 172 MBZenodo snapshot · July 26, 2026
People, groups, committees, acts, EuroVoc topics, sittings, debates and the full text of speeches.
The individual votes
N-Triples · 3.8 GBZenodo snapshot · July 26, 2026
The votes cast by each MP in every electronic ballot of the legislature, one by one.
The data derives from the open data of the Chamber of Deputies (dati.camera.it, CC BY-SA 4.0 licence), the Senate of the Republic (dati.senato.it, under its own licence) and the EuroVoc thesaurus of the EU Publications Office. Vector embeddings, user conversations and AI-generated summaries are not part of the export. GitHub · Long-term archive: Zenodo · DOI 10.5281/zenodo.21560331