A high-performance mining tool on GitHub (
https://github.com/do-me/eur-lex) for extracting text and semantic concepts from the European Commission's Cellar database.
Originally based on EuropeanParliament/Eurovoc_2025 and EuropeanParliament/Eurovoc but improved and extended for various reasons:
line breaks are not included that are quite important for semantic chunking approaches and RAG
the preprocessing was way too aggressive and removed too much… See the full description on the dataset page:
https://huggingface.co/datasets/do-me/EUR-LEX.
