Arifov's picture
Upload folder using huggingface_hub
a24ed5c verified
---
pretty_name: ECHR Case-Linked Guide Diffs
language:
- en
tags:
- legal
- echr
- temporal
- retrieval
- editing
- dataset-creation
size_categories:
- 1K<n<10K
---
# Dataset Card for ECHR Case-Linked Guide Diffs
## Dataset Summary
This dataset links case-level citation changes in ECHR guides to localized guide-body transitions, including section references and before/after paragraph text.
It is intended as a modeling-ready derived layer for tasks such as:
- guide-update retrieval
- section localization
- edit-type prediction
- constrained paragraph rewriting
## Dataset Structure
Main files:
- `case_linked_guide_diffs.csv`: case-level linked transitions
- `case_linked_guide_diffs.json`: JSON export of the same
- `case_linked_guide_diff_paragraphs.csv`: paragraph-level linked transitions
- `case_linked_guide_diff_paragraphs.json`: JSON export of the same
- `case_linked_guide_diffs_report.json`: coverage and linkage summary
## Dataset Statistics
- case-level rows: `1004`
- paragraph-level rows: `1537`
- linked case rows: `843`
- unlinked rows with existing transition pairs: `161`
- missing transition pairs: `0`
## Intended Use
This dataset is intended for:
- retrieving the guide section affected by a case
- predicting where a case should be inserted into a guide
- classifying whether an edit is an addition, removal, or revision
- constructing small-scale guide-update generation prototypes
## Provenance
Derived from:
- `lexgenie/echr-guide-citation-diffs`
- `lexgenie/echr-guide-case-catalog`
- guide-body transition parsing and paragraph alignment
This dataset should be treated as a derived body-context layer over the original citation-diff corpus.