Mavies5526/noun-courseware-sft
NOUN Courseware SFT Dataset Dataset Description This dataset contains high-quality Supervised Fine-Tuning (SFT) question-and-answer pairs generated from the National Open University of Nigeria (NOUN) course materials. It is designed to train language models on diverse academic concepts ranging from agriculture to law. Dataset Structure The dataset is divided into specific Subsets (Configs) based on the academic Faculty. You can select different… See the full description on the dataset page: https://huggingface.co/datasets/Mavies5526/noun-courseware-sft.
NOUN Courseware SFT Dataset
Dataset Description
This dataset contains high-quality Supervised Fine-Tuning (SFT) question-and-answer pairs generated from the National Open University of Nigeria (NOUN) course materials. It is designed to train language models on diverse academic concepts ranging from agriculture to law.
Dataset Structure
The dataset is divided into specific Subsets (Configs) based on the academic Faculty. You can select different faculties using the subset dropdown menu in the Dataset Viewer above.
Data Fields
- `instruction`: The generated question, prompt, or directive (e.g., "Explain cooperative legislation.").
- `response`: The comprehensive, omniscient answer extracted directly from the academic text.
- `course_file`: The source textbook the pair was generated from.
Data Cleaning & Formatting
- Meta-text Removal: Introductory rows mentioning "module", "course", or the course code were explicitly filtered out.
- Prompt Engineering: "What is..." questions were programmatically transformed into diverse action verbs (Define, Explain, Describe, Summarize, etc.) to improve the cognitive variety of the training data.