---
license: cc-by-nc-sa-4.0
language:
  - uk
tags:
  - legal
  - citation-graph
  - co-citation
  - Ukrainian
  - court-decisions
  - network-analysis
  - legal-nlp
  - arxiv:2605.15362
pretty_name: "Ukrainian Court Citation Graph (ЄДРСР)"
size_categories:
  - 1M<n<10M
task_categories:
  - graph-ml
  - text-classification
configs:
  - config_name: default
    data_files:
      - split: train
        path: cocitation_edges.parquet
dataset_info:
  - config_name: cocitation_edges
    data_files:
      - split: train
        path: cocitation_edges.parquet
    features:
      - name: law_a
        dtype: string
      - name: article_a
        dtype: string
      - name: law_b
        dtype: string
      - name: article_b
        dtype: string
      - name: weight
        dtype: int64
    splits:
      - name: train
        num_examples: 2328213
  - config_name: article_citation_stats
    data_files:
      - split: train
        path: article_citation_stats.parquet
    features:
      - name: law_number
        dtype: string
      - name: law_article
        dtype: string
      - name: citation_type
        dtype: string
      - name: total_citations
        dtype: int64
      - name: unique_decisions
        dtype: int64
    splits:
      - name: train
        num_examples: 50000
  - config_name: article_centrality
    data_files:
      - split: train
        path: article_centrality.parquet
    features:
      - name: law_number
        dtype: string
      - name: law_article
        dtype: string
      - name: degree
        dtype: int64
      - name: pagerank
        dtype: float64
      - name: authority
        dtype: float64
    splits:
      - name: train
        num_examples: 84
  - config_name: community_labels
    data_files:
      - split: train
        path: community_labels.parquet
    features:
      - name: period
        dtype: string
      - name: community_rank
        dtype: int32
      - name: community_size
        dtype: int32
      - name: dominant_law
        dtype: string
      - name: dominant_law_fraction
        dtype: float64
      - name: total_citations
        dtype: int64
    splits:
      - name: train
        num_examples: 25
---

# Ukrainian Court Citation Graph (ЄДРСР)

A large-scale co-citation graph derived from **99.5 million** Ukrainian court decisions in the Unified State Register of Court Decisions (ЄДРСР / EDRSR).

## Overview

This dataset contains the co-citation network of Ukrainian legislation as cited in court decisions from 2003 to 2026. Two legal provisions (articles) are linked if they are cited together in the same court decision. The edge weight reflects how many decisions co-cite the pair.

**Source corpus:** 99.5M full-text court decisions from [reyestr.court.gov.ua](https://reyestr.court.gov.ua)
**Citation extraction:** 345M citation rows extracted via regex patterns matching Ukrainian legal citation conventions (codex articles, law articles, constitutional provisions, case references)
**Co-citation construction:** Streaming pairwise combination of articles per decision, filtered to edges with weight >= 10

## Dataset Statistics

| Metric | Value |
|--------|-------|
| Source decisions | 99,500,000 |
| Extracted citations | 345,000,000 |
| Unique article-decision pairs | 36,958,613 |
| Co-citation edges (weight >= 10) | 2,328,213 |
| Unique articles in graph | ~50,000 |
| Time span | 2003 -- 2026 |

## Files

### `cocitation_edges.parquet` (9.4 MB)
The core co-citation graph. Each row is an undirected edge between two legal provisions.

| Column | Type | Description |
|--------|------|-------------|
| `law_a` | string | Name of the first law/codex |
| `article_a` | string | Article number in the first law |
| `law_b` | string | Name of the second law/codex |
| `article_b` | string | Article number in the second law |
| `weight` | int64 | Number of court decisions co-citing both articles |

### `article_citation_stats.parquet` (0.2 MB)
Top 50,000 most-cited legal provisions with citation counts.

| Column | Type | Description |
|--------|------|-------------|
| `law_number` | string | Law or codex name |
| `law_article` | string | Article number |
| `citation_type` | string | Type: codex_article, law_article, constitution |
| `total_citations` | int64 | Total citation count across all decisions |
| `unique_decisions` | int64 | Number of unique decisions citing this article |

### `article_centrality.parquet`
Graph centrality measures for top articles (union of top-50 by degree, PageRank, and HITS authority).

| Column | Type | Description |
|--------|------|-------------|
| `law_number` | string | Law or codex name |
| `law_article` | string | Article number |
| `degree` | int64 | Weighted degree in co-citation graph |
| `pagerank` | float64 | PageRank score |
| `authority` | float64 | HITS authority score |

### `community_labels.parquet`
Louvain community detection results across 5 time periods (2007--2010, 2011--2014, 2015--2018, 2019--2022, 2023--2026).

| Column | Type | Description |
|--------|------|-------------|
| `period` | string | Time period |
| `community_rank` | int32 | Rank by size within period |
| `community_size` | int32 | Number of articles in community |
| `dominant_law` | string | Most frequent law in the community |
| `dominant_law_fraction` | float64 | Fraction of articles from dominant law |
| `total_citations` | int64 | Sum of citations across community members |

## Key Findings

- **Power-law degree distribution** with exponent alpha = 2.17 (scale-free network)
- **PageRank diverges from raw degree:** ст. 10 ЦПК (fair trial) ranks #1 by PageRank but #2 by degree, indicating structural centrality beyond citation frequency
- **Community structure tracks legal domains:** Louvain modularity 0.54--0.61, with communities aligning to criminal, civil, administrative, and tax law
- **Temporal stability:** NMI between adjacent periods > 0.85, but post-2017 judicial reform shows community reorganization
- **Citation prediction:** Logistic regression on graph features achieves AUC = 0.9984 for predicting top-1000 articles in 2020--2026 from 2007--2019 features

## Paper

This dataset is introduced in:

> **Automatic Construction of a Legal Citation Graph from 100 Million Ukrainian Court Decisions: Large-Scale Extraction, Topological Analysis, and Ontology-Driven Clustering**
> Volodymyr Ovcharov (2026). arXiv:[2605.15362](https://arxiv.org/abs/2605.15362) [cs.CL]

## Citation

```bibtex
@article{ovcharov2026citationgraph,
  title={Automatic Construction of a Legal Citation Graph from 100 Million Ukrainian Court Decisions: Large-Scale Extraction, Topological Analysis, and Ontology-Driven Clustering},
  author={Ovcharov, Volodymyr},
  journal={arXiv preprint arXiv:2605.15362},
  year={2026},
  url={https://arxiv.org/abs/2605.15362}
}
```

## Related

- [overthelex/ua-case-outcome](https://huggingface.co/datasets/overthelex/ua-case-outcome) -- Ukrainian court case outcome prediction dataset
- [overthelex/oversight-constitution](https://huggingface.co/datasets/overthelex/oversight-constitution) -- Edit-trace oversight constitution dataset
- Paper: [Automatic Construction of a Legal Citation Graph from 100 Million Ukrainian Court Decisions](https://arxiv.org/abs/2605.15362) (arXiv:2605.15362)

## License

CC-BY-NC-SA-4.0. Academic and research use is free. Commercial use requires separate licensing from [legal.org.ua](https://legal.org.ua).