PlenaryEP: A Continuously Updated, Metadata-Rich Corpus of European Parliament Debates

We present PlenaryEP, a 109 million word corpus of debate speeches from European Parliament proceedings, together with the code to reproduce and update it with the newest proceedings releases. The corpus spans the period from 1996 to 2026 and has been extensively cleaned, filtered, and translated into English. The speeches are enriched with metadata of the Members of the European Parliament (MEP), linked to the party stances of the Chapel Hill Expert Survey (CHES), and automatically annotated with the policy areas of the Comparative Agendas Project (CAP). We verified that all forms of automatic and semi-supervised annotation are effective. Our extraction, preprocessing, and enrichment framework is scalable and allows reproducing or updating the corpus on limited hardware.