LegalPincite is a large-scale legal IR dataset derived from CJEU judgments that includes masked case and paragraph-level queries. It addresses data leakage and oversimplification in existing datasets by providing realistic, paragraph-specific citation annotations.
HOW THIS AFFECTS YOU
●
builderYou can train and evaluate more robust legal retrieval systems using realistic paragraph-level data.
●
researcherYou can use this dataset to study more complex and realistic legal information retrieval scenarios.