---
license: other
language:
- ru
task_categories:
- token-classification
pretty_name: Russian Legal NER
tags:
- legal
- ner
- russian
- court-decisions
---

# Russian Legal NER

This dataset contains Russian legal/court decision texts with span-level NER annotations.

Original source: [Google Drive folder](https://drive.google.com/drive/folders/1g5H3LsgZnYdPhtZntYFI4eUzbnzdXULf).

All rights and authorship belong to the original dataset authors. This Hugging Face mirror is provided for easier reproducible training and evaluation.

## Format

Files are provided as Parquet splits:

- `data/train.parquet`
- `data/validation.parquet`
- `data/test.parquet`

Columns:

- `text`: document text.
- `spans`: string representation of span annotations: `[start, end, surface, label_id]`.

Observed label ids:

- `2`
- `4`
- `9`
- `13`
- `17`

## Loading

```python
from datasets import load_dataset

ds = load_dataset("TryDotAtwo/russian-legal-ner")
```