An Expanded Massive Multilingual Dataset for High-Performance Language Technologies (HPLT)
: Burchell, Laurie; De Gibert Bonet, Ona; Arefyev, Nikolay; Aulamo, Mikko; Bañón, Marta; Chen, Pinzhen; Fedorova, Mariia; Guillou, Liane; Haddow, Barry; Hajič, Jan; Helcl, Jindřich; Henriksson, Erik; Klimaszewski, Mateusz; Komulainen, Ville; Kutuzov, Andrey; Kytöniemi, Joona; Laippala, Veronika; Mæhlum, Petter; Malik, Bhavitvya; Mehryary, Farrokh; Mikhailov, Vladislav; Moghe, Nikita; Myntti, Amanda; O’Brien, Dayyán; Oepen, Stephan; Pal, Proyag; Piha, Jousia; Pyysalo, Sampo; Ramírez-Sánchez, Gema; Samuel, David; Stepachev, Pavel; Tiedemann, Jörg; Variš, Dušan; Vojtěchová, Tereza; Zaragoza-Bernabeu, Jaume
: Che, Wanxiang; Nabende, Joyce; Shutova, Ekaterina; Pilehvar, Mohammad Taher
: Annual Meeting of the Association for Computational Linguistics
Publisher: Association for Computational Linguistics
: 2025
Annual Meeting of the Association for Computational Linguistics
: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
: 17452
: 17485
: 0736-587X
DOI: https://doi.org/10.18653/v1/2025.acl-long.854
: https://doi.org/10.18653/v1/2025.acl-long.854
: https://research.utu.fi/converis/portal/detail/Publication/505515303
Training state-of-the-art large language models requires vast amounts of clean and diverse textual data. However, building suitable multilingual datasets remains a challenge. In this work, we present HPLT v2, a collection of high-quality multilingual monolingual and parallel corpora, extending prior work of the HPLT project. The monolingual portion of the data contains 8T tokens covering 193 languages, while the parallel data contains 380M sentence pairs covering 51 languages. We document the entire data pipeline and release the code to reproduce it. We provide extensive analysis of the quality and characteristics of our data. Finally, we evaluate the performance of language models and machine translation systems trained on HPLT v2, demonstrating its value.
:
This project has received funding from the European Union’s Horizon Europe research and innovation programme under Grant agreement No 101070350 and from UK Research and Innovation (UKRI) under the UK government’s Horizon Europe funding guarantee [grant number 10052546].
