CoolFace
Datasetpublic

nassimjp/pashto-math-questions

Pashto Math Questions Dataset This dataset contains a clean collection of mathematical word problems translated into Pashto and localized for regional context. The original source data consisted of raw conversational Chinese math strings. Because many machine-translated math solutions contain logical and arithmetic inaccuracies, this dataset extracts and preserves only the question fields to provide a high-quality foundation for custom mathematical alignment, instruction tuning… See the full description on the dataset page: https://huggingface.co/datasets/nassimjp/pashto-math-questions.

sourceHugging Facemitupdated 3mo agoView on Hugging Face
0likes11downloads
Dataset Card

Pashto Math Questions Dataset

This dataset contains a clean collection of mathematical word problems translated into Pashto and localized for regional context.

The original source data consisted of raw conversational Chinese math strings. Because many machine-translated math solutions contain logical and arithmetic inaccuracies, this dataset extracts and preserves only the question fields to provide a high-quality foundation for custom mathematical alignment, instruction tuning, or fine-tuning evaluation.

Dataset Structure

The dataset follows the standard OpenAI/Hugging Face messages conversational schema configuration, leaving the assistant role empty for targeted fine-tuning hooks:

json
{
  "messages": [
    {
      "role": "user",
      "content": "پوښتنه: خان هر سهار ښوونځي ته په تګ ۱۰ دقیقې وخت تېروي. که د خان کور له ښوونځي څخه 2 کیلومتره لرې وي، هغه په ​​یوه دقیقه کې څو متره مزل کوي؟"
    },
    {
      "role": "assistant",
      "content": ""
    }
  ]
}

Features & Enhancements

  • —Question-Only Target: Excludes flawed chain-of-thought steps from base machine translations, providing pristine prompts for modern LLM evaluation or response generation.
  • —Cultural Localization: Common Chinese Pinyin placeholders (e.g., Xiaoming, Xiaohua) have been systematically replaced with common Pashto names (e.g., Khan, Malala) during processing to preserve natural syntactic flows.
  • —Math Vocabulary Patching: Fixes common machine translation literal bugs (such as shifting literal adverbial translations of "even" over to correct mathematical Pashto terminology like "جفت شمیر").