Alignement Bilingue des Résumés et des Mots-Clés de theses.fr

English–French parallel corpus extracted from defended doctoral theses available through theses.fr. Details of the data collection and corpus construction are provided in: https://openreview.net/forum?id=BGcECDvEjw The scripts used to collect and process the corpus are available at:https://github.com/ANR-MaTOS/Resources/tree/main/theses-fr-ARTS2026 It consists of aligned.theses.fr.with-cometkiwi-year-discipline.parquet English--French parallel abstracts, aligned at sentence level, each record includes the docid, COMETKiwi scores, year and discipline. theses.fr.titles.en-fr.with-cometkiwi-lid-prob.parquetEnglish--French parallel thesis titles, with COMETKiwi scores and the language identification probabilities for both title variants. theses.fr.en-fr-keywords.clean-lid.grouby-doc.parquetEnglish--French parallel keywords extracted from thesis metadata and grouped by thesis document. keywords_by_en.json.gzEnglish–French keyword pairs indexed by the English keyword, facilitating lookup and analysis from the English side of the corpus. keywords_by_fr.json.gzEnglish–French keyword pairs indexed by the French keyword. to cite @inproceedings{ peng-etal-2026-alignment, title={Alignement Bilingue des R\'esum\'es et des Mots-Cl\'es de theses.fr}, author={Ziqian Peng, Lichao Zhu, Maxime Bouthors, François Yvon}, booktitle={Atelier sur l'Analyse et la Recherche de Textes Scientifiques}, year={2026}, url={https://openreview.net/forum?id=BGcECDvEjw}, note={to appear at ARTS 2026} }