nassimjp/food-safety-classifications-pashto
🥗 Food‑Safety‑Classifications‑Pashto Pashto Dataset for Food‑Safety Statement Classification & SFT Conversation Generation 📌 Overview Food‑Safety‑Classifications‑Pashto یو Pashto ژبنی ډیټاسټ دی چې پکې د خوراکي خوندیتوب په اړه جملې شاملې دي، او هره جمله د صحت/خوندیتوب د اعتبار له مخې طبقه بندي شوې ده. هره کرښه په اصل کې داسې ده: {"text": "...", "label": 0} {"text": "...", "label": 1} label = 0 → جمله سمه ده (Correct) label = 1 → جمله ناسمه ده (Incorrect) دا… See the full description on the dataset page: https://huggingface.co/datasets/nassimjp/food-safety-classifications-pashto.
🥗 Food‑Safety‑Classifications‑Pashto
Pashto Dataset for Food‑Safety Statement Classification & SFT Conversation Generation
📌 Overview
Food‑Safety‑Classifications‑Pashto یو Pashto ژبنی ډیټاسټ دی چې پکې د خوراکي خوندیتوب په اړه جملې شاملې دي، او هره جمله د صحت/خوندیتوب د اعتبار له مخې طبقه بندي شوې ده.
هره کرښه په اصل کې داسې ده:
{"text": "...", "label": 0}
{"text": "...", "label": 1}
- label = 0 → جمله سمه ده (Correct)
- label = 1 → جمله ناسمه ده (Incorrect)
دا ډیټاسټ د classification، safety alignment، او SFT conversation generation لپاره کارول کېږي.
🎯 Purpose
دا ډیټاسټ د لاندې کارونو لپاره جوړ شوی:
- د Pashto ژبې لپاره food‑safety classification models روزل
- د LLM safety alignment لپاره د غلطو/سمو جملو تشخیص
- د SFT supervised fine‑tuning لپاره د مکالموي بڼې تولید
- د Pashto ژبې لپاره د reasoning datasets پراخول
- د Talanda او نورو Pashto‑LLM پروژو لپاره د پاکو، معیاري ډیټاسټونو جوړول
📂 Dataset Structure
🔸 Original Format (JSONL)
هره کرښه داسې ده:
{"text": "It's okay to leave leftovers at room temperature for up to 2 hours before refrigerating them.", "label": 0}
🔸 SFT Conversation Format
د supervised fine‑tuning لپاره د مکالمې بڼه:
{
"id": "000000",
"messages": [
{"role": "user", "content": "دا سمه ده چې پاتې شوي توکي د یخچال کولو دمخه تر 2 ساعتونو پورې د خونې په حرارت کې پریږدئ."},
{"role": "assistant", "content": "هو، دا بیان سم دی."}
]
}
🧪 Example Entries
✔️ Correct Statement (label = 0)
{"text": "Leftovers can be safely kept at room temperature for up to 2 hours.", "label": 0}
❌ Incorrect Statement (label = 1)
{"text": "Raw chicken can be handled with the same utensils used for salads.", "label": 1}
🔧 Conversion Script (JSONL → SFT Conversations)
د Pashto SFT بڼه تولیدولو لپاره سکریپټ:
import json
input_path = "input.jsonl"
output_path = "sft_conversations.jsonl"
def convert_to_sft(text, label):
if label == 0:
messages = [
{"role": "user", "content": text},
{"role": "assistant", "content": "هو، دا بیان سم دی."}
]
else:
messages = [
{"role": "user", "content": text},
{"role": "assistant", "content": "نه، دا بیان ناسم دی."}
]
return {"messages": messages}
with open(input_path, "r", encoding="utf-8") as fin, \
open(output_path, "w", encoding="utf-8") as fout:
for line in fin:
if not line.strip():
continue
obj = json.loads(line)
sft_obj = convert_to_sft(obj["text"], obj["label"])
fout.write(json.dumps(sft_obj, ensure_ascii=False) + "\n")
📊 Dataset Statistics
🧰 Use Cases
- Food Safety Classification Models
- LLM Safety Alignment
- Pashto Reasoning Tasks
- Supervised Fine‑Tuning (SFT)
- Binary Classification Training
- Dataset Augmentation for Pashto NLP
📥 Loading the Dataset
from datasets import load_dataset
ds = load_dataset("nassimjp/food-safety-classifications-pashto")
print(ds["train"][0])
📜 License
This dataset is released under the MIT License.
You may use, modify, and distribute it freely with attribution.
🤝 Contributions
که غواړې چې نور Pashto food‑safety جملې، اصلاحات، یا مکالمې اضافه کړې، نو د Pull Request له لارې یې شریکولای شې.
🔗 Related Work
- Pashto SFT Dataset
- Pashto Safety Alignment
- Talanda NLP Framework
