G4KMU/LEMUR

EU Law Dataset – Category 15.10: Environment

This dataset contains official legal documents from the European Union, collected from the EUR-Lex website, specifically under category 15.10: "Environment". The documents span from the year 1961 to 2025 and are provided in multiple European "languages. The original documents are in PDF format and have been converted into various text-based formats using OLMCR. The dataset splits represent the different "languages available for each document. It is designed for multilingual legal document retrieval. Queries consist of structured metadata, and each query is paired with the corresponding legal document as a positive retrieval example.

🔗 Links

• Paper: https://arxiv.org/abs/2602.09570
• Code: https://github.com/nargesbh/eur_lex

Dataset Structure

Each document represents a legal act and is available in several European "languages. The dataset includes the same set of documents processed through different document conversion methods, organized into pages represented as rows in the dataset. Each row contains the following fields:

• law_id: Unique identifier of the legal document.
• category: Category identifier of the legal document (e.g., category15).
• year: Year the legal document was published.
• primary_language: Language code of the document (e.g., de).
• metadata: Document header and front-matter text extracted from the source.
• text: Full document text content, excluding metadata.
• is_table: Boolean indicating whether any page in the document contains a table.
• is_diagram: Boolean indicating whether any page in the document contains a diagram.
• rotation_correction: Maximum absolute rotation correction (in degrees) applied across document pages.
• is_rotation_valid: Boolean indicating whether rotation correction was valid for all pages.
• pdf_path: Relative path to the original PDF file of the document.
• pdf_total_pages: Total number of pages in the original PDF document.

languages

The dataset covers the 23 official languages of the European Union:

• Bulgarian (BG)
• Croatian (HR)
• Czech (CS)
• Dutch (NL)
• English (EN)
• Estonian (ET)
• Finnish (FI)
• French (FR)
• German (DE)
• Greek (EL)
• Hungarian (HU)
• Irish (GA)
• Italian (IT)
• Latvian (LV)
• Lithuanian (LT)
• Maltese (MT)
• Polish (PL)
• Portuguese (PT)
• Romanian (RO)
• Slovak (SK)
• Slovenian (SL)
• Spanish (ES)
• Swedish (SV)

Older documents may not include all 23 languages due to EU membership timelines, but more recent documents are consistently available in all languages.

Citation Information

@inproceedings{ahmadi-etal-2026-lemur,
    title = "{LEMUR}: A Corpus for Robust Fine-Tuning of Multilingual Law Embedding Models for Retrieval",
    author = "Ahmadi, Narges Baba  and
      Strich, Jan  and
      Semmann, Martin  and
      Biemann, Chris",
    editor = "Baez Santamaria, Selene  and
      Somayajula, Sai Ashish  and
      Yamaguchi, Atsuki",
    booktitle = "Proceedings of the 19th Conference of the {E}uropean Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 4: Student Research Workshop)",
    month = mar,
    year = "2026",
    address = "Rabat, Morocco",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.eacl-srw.18/",
    doi = "10.18653/v1/2026.eacl-srw.18",
    pages = "248--265",
    ISBN = "979-8-89176-383-8",
    abstract = "Large language models (LLMs) are increasingly used to access legal information. Yet, their deployment in multilingual legal settings is constrained by unreliable retrieval and the lack of domain-adapted, open-embedding models. In particular, existing multilingual legal corpora are not designed for semantic retrieval, and PDF-based legislative sources introduce substantial noise due to imperfect text extraction. To address these challenges, we introduce LEMUR, a large-scale multilingual corpus of EU environmental legislation constructed from 24,953 official EUR-Lex PDF documents covering 25 languages. We further propose the Lexical Content Score (LCS), a language-agnostic metric that quantifies the fidelity of PDF-to-text conversion by measuring lexical consistency against authoritative HTML versions. Building on LEMUR, we fine-tune three state-of-the-art multilingual embedding models using contrastive objectives in both monolingual and bilingual settings, reflecting realistic legal-retrieval scenarios. Experiments across low- and high-resource languages demonstrate that legal-domain fine-tuning consistently improves Top-k retrieval accuracy relative to strong baselines, with particularly pronounced gains for low-resource languages. Cross-lingual evaluations show that these improvements transfer to unseen languages, indicating that fine-tuning primarily enhances language-independent, content-level legal representations rather than language-specific cues. We publish code[GitHub Repository] and data[Hugging Face Dataset]."
}

Source and Licensing

• Source: EUR-Lex – Access to European Union Law
• Category: 15.10 Environment (under directory 15: Environment, Consumers and Health Protection)
• Website: https://eur-lex.europa.eu/browse/directories/legislation.html
• Years Covered: 1961 to 2025