Skip to content

The research the platform is built on

EURVISTA is a translational project. The methods it applies to the EU legislative record, tracing amendments through text reuse, tracking parliamentary attention over time, reading stakeholder influence from the committee record, were developed and peer reviewed first. The platform is what happens when that work is turned into something usable by people outside a university.

01 / Questions

What the project is trying to answer

Can a machine trace influence through a legislative record?

Amendments, meetings and votes leave traces. The question is whether those traces can be joined reliably enough across six registers and three institutions to say something defensible about who moved a file, and where the joins break down.

Where do language models distort political speech?

Summarisation is where an AI system is most useful and most dangerous. Our work measures where models systematically misrepresent what was argued, and which speakers lose out when they do. That finding is why generated prose on the platform is marked rather than presented as fact.

What does honest uncertainty look like in a policy tool?

A tool used to brief decision-makers has to be legible about what it does not know. Ordinal bands instead of scores, stated coverage instead of implied completeness, historical frequencies instead of forecasts: these are research positions as much as design ones.

02 / Publications

EU legislative politics and transparency

2017 · EUP
Cross & Hermansson, European Union Politics. Sage Prize for best article. The text-reuse method behind the platform’s amendment tracing.
2021 · WEP
Cross, Eising, Hermansson & Spohr, West European Politics. How stakeholder input shapes committee amendments, the relationship the stakeholder map is built around.
2013 · EJPR
Cross, European Journal of Political Research. How publicity changes recorded national positions, background to the Council views.
2020 · EUP
Cross & Vaznonytė, European Union Politics. Legislative timing and throughput, the ground the duration base rates stand on.
2012 · JEPP
Thomson, Arregui, Leuffen, Costello, Cross, Hertz & Jensen, Journal of European Public Policy.
03 / Publications

Text-as-data and language models for parliaments

2017 · Political Analysis
Greene & Cross. Tracking parliamentary attention over time from the record, the approach behind the platform’s attention signals.
2026 · KR
Evaluating LLM-driven summarisation of parliamentary debates with computational argumentation
Cunningham, Greene, Cross & Rago, KR 2026, KR in the Wild track. How to check that AI summaries of debates stay faithful to what was actually argued.
2026 · ECML PKDD
Cunningham, Cross & Greene. Where summarisation models systematically distort political speech, and how to detect it.
2025 · CCR

Full publication record →

04 / Funding

Who pays for this

EURVISTA is funded by the All-Ireland Research Council’s Translational Research Programme, which supports moving proven academic research into deployed products. It is hosted at University College Dublin.

The machine learning and natural language processing side of the work is done with Rinn AI, the Research Ireland Centre for Research and Innovation in Data Science and AI, which succeeded the Insight Centre for Data Analytics in July 2026.

EURVISTA holds no third-party licensed data. Every record is drawn from a public institutional register.

Who is building it →