Democracy & eGov · Horizon Europe · 2022–2025
Augmenting deliberative democracy at every scale.
Novelcore bridges Horizon Europe research and production-grade AI — building deliberation, knowledge and infrastructure tools for the public sector and beyond.
consortium
10 partners, 6 countries
our role
Technical Coordinator
programme
horizon europe
Grant reference
101094665
Our role
Technical Coordinator
Novelcore leads the technical work of the ORBIS consortium — architecting the deliberation platform's argumentation knowledge graphs and NLP pipeline, and coordinating delivery across ten partner institutions. ORBIS is the research testbed that feeds directly into dlbrt., our production deliberation product.
Impact and outcomes
From research prototype to deployable pattern.
ORBIS validated an approach that now underpins dlbrt., proving out the deliberation model with real institutional partners before it reached production.
Presented at the Athens Democracy Forum in partnership with the New York Times, reaching an audience of policy and civic-tech practitioners.
Argumentation graph model validated across pilot deliberations run with consortium partners.
Technical architecture and lessons learned carried forward directly into dlbrt.'s production platform.
The challenge
Public opinion is everywhere. Structure is nowhere.
Every year, institutions run consultations, town halls, and online forums that generate thousands of comments — and almost none of it gets synthesized into something a policymaker can act on. The result: citizens feel unheard, and decision-makers miss signal buried in noise.
ORBIS set out to close that gap by building a platform that doesn't just collect opinions, but understands their structure — arguments, counter-arguments, areas of consensus and genuine disagreement — at a scale no human moderation team could match.
our approach
Knowledge graphs that argue back.
We designed and built the project's core knowledge representation layer: an argumentation graph model that captures claims, evidence, and the relationships between them, extracted automatically from unstructured text using our NLP pipeline.
On top of that, we coordinated the integration work across partners — connecting the extraction models, the visualization layer, and the citizen-facing interface into a single working pipeline, and led the technical work packages through three consortium review cycles.