omerfaksoy's picture
Update README.md
e1d489d verified
|
Raw
History Blame Contribute Delete
5.91 kB
---
language:
- tr
license: cc-by-4.0
task_categories:
- text-generation
- text-classification
- token-classification
tags:
- turkish
- recipe
- cooking
- instruction-tuning
- nlp
- food
pretty_name: Turkish Recipe Corpus (TRC-30K)
size_categories:
- 10K<n<100K
configs:
- config_name: structured
data_files:
- split: train
path: structured/train-*
- split: validation
path: structured/validation-*
- split: test
path: structured/test-*
- config_name: instruction
data_files:
- split: train
path: instruction/train-*
- split: validation
path: instruction/validation-*
- split: test
path: instruction/test-*
- config_name: ingredient2recipe
data_files:
- split: train
path: ingredient2recipe/train-*
- split: validation
path: ingredient2recipe/validation-*
- split: test
path: ingredient2recipe/test-*
---
# Turkish Recipe Corpus (TRC-30K)
**Türkçe'nin en kapsamlı açık kaynaklı tarif veri seti.**
74,768 Türkçe tariften oluşan, 3 farklı NLP görevine hazır yapılandırılmış corpus.
> Ethosoft Research · [huggingface.co/Ethosoft](https://huggingface.co/Ethosoft) · [ethosoft.org](https://ethosoft.org)
---
## Dataset Özeti
| | Değer |
|---|---|
| **Toplam tarif** | 74,768 |
| **Dil** | Türkçe (tr) |
| **Lisans** | CC BY 4.0 |
| **Konfigürasyonlar** | 3 (`structured`, `instruction`, `ingredient2recipe`) |
| **Split** | Train / Validation / Test (80 / 10 / 10) |
| **Ortalama adım sayısı** | 5.23 / tarif |
| **Ortalama malzeme sayısı** | 6.85 / tarif |
---
## Konfigürasyonlar
### `structured` - Yapılandırılmış Tarif Verisi
Parse edilmiş, kategorize edilmiş ham tarif verisi. NER, sınıflandırma ve
veri analizi görevleri için idealdir.
```python
from datasets import load_dataset
ds = load_dataset("Ethosoft/Turkish-Recipe-Corpus", "structured")
print(ds["train"][0])
```
**Alan şeması:**
| Alan | Tür | Açıklama |
|---|---|---|
| `recipe_name` | string | Temizlenmiş yemek adı |
| `ingredients` | list[string] | Parse edilmiş malzeme listesi |
| `steps` | list[string] | Numaralı yapılış adımları |
| `category` | string | Yemek kategorisi |
| `ingredient_count` | int | Malzeme sayısı |
| `step_count` | int | Adım sayısı |
| `raw_ingredients` | string | Ham malzeme metni |
| `raw_steps` | string | Ham tarif metni |
### `instruction` - Türkçe Instruction Tuning Verisi
Türkçe LLM'leri fine-tune etmek için chat formatında hazırlanmış tarif örnekleri.
Türkçe SFT (Supervised Fine-Tuning) için doğrudan kullanılabilir.
```python
from datasets import load_dataset
ds = load_dataset("Ethosoft/Turkish-Recipe-Corpus", "instruction")
print(ds["train"][0])
```
**Alan şeması:**
| Alan | Tür | Açıklama |
|---|---|---|
| `system` | string | Sistem mesajı |
| `user` | string | Kullanıcı sorusu (5 farklı şablondan random) |
| `assistant` | string | Markdown formatında tarif cevabı |
| `recipe_name` | string | Yemek adı |
| `category` | string | Kategori |
### `ingredient2recipe` - Malzemeden Tarif Üretimi
Verilen malzeme listesinden tam tarif üretme görevi için seq2seq / generation
benchmark'ı. Kendi modelinizi değerlendirmek için kullanın.
```python
from datasets import load_dataset
ds = load_dataset("Ethosoft/Turkish-Recipe-Corpus", "ingredient2recipe")
print(ds["train"][0])
```
**Alan şeması:**
| Alan | Tür | Açıklama |
|---|---|---|
| `input` | string | Virgülle ayrılmış malzeme listesi |
| `output` | string | Tam tarif (markdown) |
| `recipe_name` | string | Yemek adı |
| `category` | string | Kategori |
---
## Kategori Dağılımı
| Kategori | Tarif Sayısı |
|---|---:|
| tatlı & pasta | 21760 |
| diğer | 20204 |
| börek & hamur | 9747 |
| et yemekleri | 5693 |
| salata & meze | 4039 |
| sebze & zeytinyağlı | 3259 |
| çorba | 3182 |
| tavuk & kümes | 3128 |
| pilav & tahıl | 1271 |
| kahvaltılık | 1144 |
| balık & deniz | 971 |
| içecek & şurup | 370 |
---
## Kullanım Alanları
- **Türkçe LLM Fine-Tuning** - `instruction` config ile SFT
- **Yemek Kategorisi Sınıflandırma** - `structured` config ile text classification
- **Türkçe NER** - Malzeme adı, ölçü birimi, miktar entity extraction
- **Generation Benchmark** - `ingredient2recipe` ile seq2seq değerlendirmesi
- **Türk Mutfağı Analizi** - Malzeme kullanım frekansı ve örüntü analizi
---
## İlgili Kaynaklar
| Kaynak | Link |
|---|---|
| NedoTurkishTokenizer | [Ethosoft/NedoTurkishTokenizer](https://huggingface.co/Ethosoft/NedoTurkishTokenizer) |
| OpenR1-Math-220k-Turkish | [Ethosoft/OpenR1-Math-220k-Turkish](https://huggingface.co/datasets/Ethosoft/OpenR1-Math-220k-Turkish) |
| Ethosoft HuggingFace | [huggingface.co/Ethosoft](https://huggingface.co/Ethosoft) |
| Ethosoft Web | [ethosoft.org](https://ethosoft.org) |
---
## Lisans
Bu veri seti **Creative Commons Attribution 4.0 International (CC BY 4.0)** lisansı
altında yayımlanmıştır. Atıf yapılması koşuluyla ticari ve akademik kullanıma açıktır.
---
## Atıf
```bibtex
@dataset{ethosoft2025trc,
title = {Turkish Recipe Corpus (TRC-30K)},
author = {Aksoy, {\"O}mer Faruk and Dalar, Ya{\u{g}}{\i}z Ekrem and Saliho{\u{g}}lu, Feyzi Arda and
Sezer, Nedim Mutlu and {\"O}zt{\"u}rk, Ahmet R{\i}fat},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/Ethosoft/Turkish-Recipe-Corpus},
note = {Ethosoft Research, Pertevniyal High School, Istanbul}
}
```
---
*Ethosoft Research - Pertevniyal Lisesi, İstanbul, Türkiye*