README.md

---
license: cc-by-nc-4.0
task_categories:
- token-classification
task_ids:
- named-entity-recognition
language:
- tr
tags:
- legal
- legal-ner
- turkish-legal
- turkish-ner
- turkish-legal-ner
- turkish-nlp
pretty_name: TLNER
size_categories:
- 1K<n<10K
dataset_info:
  features:
  - name: tokens
    list: string
  - name: ner_tags
    list:
      class_label:
        names:
          '0': B-CA
          '1': B-COU
          '2': B-DATE
          '3': B-DEC
          '4': B-LEG
          '5': B-PER
          '6': B-ROLE
          '7': I-CA
          '8': I-COU
          '9': I-DATE
          '10': I-DEC
          '11': I-LEG
          '12': I-PER
          '13': I-ROLE
          '14': O
  splits:
  - name: train
    num_bytes: 1484239
    num_examples: 1509
  - name: validation
    num_bytes: 183537
    num_examples: 189
  - name: test
    num_bytes: 178675
    num_examples: 189
  download_size: 307498
  dataset_size: 1846451
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
source_datasets:
- original
---
# TLNER: Turkish Legal Named Entity Recognition Dataset ⚖️ 🇹🇷

TLNER is a densely annotated, domain-specific dataset specifically designed for Named Entity Recognition (NER) tasks in Turkish judicial texts. It contains formal court decisions derived from the Council of State (Danıştay) and the Court of Cassation (Yargıtay), the highest judicial bodies in Turkey. For more details about the dataset, methodology, and experiments, you can refer to the corresponding [research paper](https://link.springer.com/article/10.1007/s44443-026-00915-z).

---
## Citation
If you use this dataset, please cite the following paper:

```
@article{incidelen2026workflow,
  title={A workflow-oriented and risk-aware system for Turkish legal named entity recognition: integrating transformer-based models with legal knowledge graphs},
  author={{\.I}ncidelen, Mert and Aydo{\u{g}}an, Murat},
  journal={Journal of King Saud University Computer and Information Sciences},
  year={2026},
  publisher={Springer}
  doi={10.1007/s44443-026-00915-z}
}
```
---
## Dataset Overview
- **Number of Sentences**: 1,887
- **Number of Total Tokens**: 100,867
- **Number of Entity Tokens**: 32,920
- **Entity Density**: 32.64%
- **Languages**: Turkish

### Dataset Structure
The dataset is divided into three subsets for training, validation, and testing:

| Split      | Number of Sentences | Number of Tokens |
|------------|---------------------|-------------------|
| Training   | 1,509               | 80,976            |
| Validation | 189                 | 10,060            |
| Testing    | 189                 | 9,831             |

### Entity Taxonomy
The dataset includes 7 domain-specific legal entity types annotated using the standard BIO scheme:
- **CA** (Case Number): Unique identifiers for judicial files, including both case and decision numbers.
- **COU** (Court): Judicial bodies, chambers, and specific courts.
- **DATE** (Date): Temporal expressions.
- **DEC** (Judicial Decision): Formal judicial outcomes and verdict expressions.
- **LEG** (Legislation): Statutory references like laws and articles.
- **PER** (Person): Names of individuals mentioned in the text.
- **ROLE** (Legal Role): Institutional roles and titles of the parties in the legal context.

---
## How to Use

This dataset can be used with libraries such as [🤗 Datasets](https://huggingface.co/docs/datasets) or [pandas](https://pandas.pydata.org/). Below are examples of the use of the dataset:

```python
from datasets import load_dataset

dataset = load_dataset("incidelen/TLNER")

train_data = dataset["train"]
val_data = dataset["validation"]
test_data = dataset["test"]
```