CraneAILabs/pedagogy-benchmark-nyankore
收藏Hugging Face2025-10-28 更新2026-01-03 收录
下载链接:
https://hf-mirror.com/datasets/CraneAILabs/pedagogy-benchmark-nyankore
下载链接
链接失效反馈官方服务:
资源简介:
---
license: apache-2.0
task_categories:
- question-answering
- multiple-choice
language:
- nyn # Nyankore
pretty_name: Pedagogy Benchmark - Nyankore
size_categories:
- 1K<n<10K
tags:
- education
- pedagogy
- multilingual
- african-languages
- teacher-training
configs:
- config_name: nyankore
data_files:
- split: cdpk_main
path: nyankore/cdpk_main-*.parquet
- split: cdpk_send
path: nyankore/cdpk_send-*.parquet
---
# Pedagogy Benchmark - Nyankore
A Nyankore translation of the [AI-for-Education/pedagogy-benchmark](https://huggingface.co/datasets/AI-for-Education/pedagogy-benchmark) dataset.
## Dataset Description
This dataset provides translations of Chilean teacher training exam questions into Nyankore (Uganda). The original dataset contains multiple-choice questions covering various pedagogical domains, education levels, and subject areas.
### Languages
- **Nyankore (Uganda)** - nyn
### Translation Methodology
All translations were performed using **Sunbird/Sunflower-32B** model with:
- Specialized system prompts for educational content
- Retry logic with exponential backoff for reliability
- Human validation on sample sets
- Preservation of all metadata and structure
- 100% clean translations (no English contamination)
## Dataset Structure
### Configs and Splits
The dataset has two splits:
- **`cdpk_main`**: Main pedagogy questions (920 questions)
- **`cdpk_send`**: Special Educational Needs and Disabilities (SEND) questions (223 questions)
### Data Fields
- `question_id` (int): Unique identifier
- `question` (string): Multiple-choice question text
- `answer_a`, `answer_b`, `answer_c`, `answer_d` (string): Answer options
- `answer_e`, `answer_f`, `answer_g` (string, nullable): Additional answer options
- `correct_answer` (string): Correct answer letter (A-G)
- `category` (string): Subject category (e.g., Science, SEND, General)
- `pedagogical_subdomain` (string): Pedagogical area (e.g., Assessment, Teaching strategies)
- `age_group` (string): Educational level (Pre-primary, Primary, Secondary)
- `year` (int): Exam year
- `secondary_category` (string, nullable): Additional categorization for SEND questions
### Data Splits
| Language | cdpk_main | cdpk_send | Total |
|----------|-----------|-----------|-------|
| Nyankore | 920 | 223 | 1,143 |
## Usage
### Load a specific split
```python
from datasets import load_dataset
# Load Nyankore main questions
dataset = load_dataset("CraneAILabs/pedagogy-benchmark-nyankore", "nyankore", split="cdpk_main")
# Load Nyankore SEND questions
dataset = load_dataset("CraneAILabs/pedagogy-benchmark-nyankore", "nyankore", split="cdpk_send")
# Access a question
print(dataset[0]['question'])
print(f"A) {dataset[0]['answer_a']}")
print(f"B) {dataset[0]['answer_b']}")
print(f"C) {dataset[0]['answer_c']}")
print(f"D) {dataset[0]['answer_d']}")
print(f"Correct: {dataset[0]['correct_answer']}")
```
### Load all splits
```python
from datasets import load_dataset
# Load both splits
dataset_dict = load_dataset("CraneAILabs/pedagogy-benchmark-nyankore", "nyankore")
print(f"Main questions: {len(dataset_dict['cdpk_main'])}")
print(f"SEND questions: {len(dataset_dict['cdpk_send'])}")
```
## Dataset Statistics
### Question Categories (cdpk_main)
The main split covers various subject areas:
- Science
- Language and Literature
- Mathematics
- Social Studies
- Arts and Physical Education
- General Pedagogy
### Pedagogical Subdomains
Questions assess knowledge across:
- Assessment and evaluation
- Teaching strategies
- Student understanding
- Curriculum and planning
- Education theories
- Professional development
### Education Levels
- **Pre-primary**: Early childhood education
- **Primary**: Elementary education
- **Secondary**: High school education
### SEND Questions (cdpk_send)
The SEND split focuses on:
- Special educational needs
- Inclusive education strategies
- Differentiated instruction
- Individual education plans
- Assistive technologies
## Original Dataset
This is a translated version of:
**Pedagogy Benchmark**
Original Dataset: [AI-for-Education/pedagogy-benchmark](https://huggingface.co/datasets/AI-for-Education/pedagogy-benchmark)
**Citation:**
```bibtex
@misc{pedagogy-benchmark-2024,
title={Pedagogy Benchmark: Chilean Teacher Training Exams},
author={AI for Education},
year={2024},
publisher={Hugging Face},
url={https://huggingface.co/datasets/AI-for-Education/pedagogy-benchmark}
}
```
## Translation Details
**Translated by:** CraneAI Labs
**Translation Date:** October 2025
**Model:** Sunbird/Sunflower-32B
**Language:** Nyankore
### Quality Assurance
- Educational terminology verified by native speakers
- Sample validation across all question types
- Metadata and structure preserved exactly
- Checkpoint-based translation with error recovery
- Retry logic ensures 100% clean translations (no English fallbacks)
## Limitations
- Automated translations may not capture all cultural nuances
- Some pedagogical terms may have multiple valid translations
- Context-specific educational references may need adaptation
- Recommended for use with human review for high-stakes applications
## License
This dataset maintains the **Apache 2.0** license from the original pedagogy-benchmark dataset.
## Contact
For questions or issues with translations:
- **Organization:** [CraneAI Labs](https://huggingface.co/CraneAILabs)
- **Original Dataset:** [AI-for-Education](https://huggingface.co/AI-for-Education)
## Acknowledgments
- Original dataset creators at AI-for-Education
- Chilean Ministry of Education (MINEDUC) for source materials
- Sunbird AI for the Sunflower-32B translation model
- Nyankore language speakers who validated translations
提供机构:
CraneAILabs


