language:
- hi
license: mit
task_categories:
- text-generation
Hindi Simple Stories
Overview
Hindi Simple Stories is a Hindi translation of the Simple Stories dataset. It is designed as a clean corpus of simple, coherent stories for training and evaluating Hindi language models.
The dataset preserves the structure and meaning of the original stories while making them available in Hindi for language modeling research.
Motivation
Despite rapid progress in large language models, there are relatively few high-quality, openly available Hindi datasets specifically designed for training small language models on coherent narrative text.
This dataset aims to provide a simple, high-quality corpus suitable for:
- Hindi language model pretraining
- Tokenizer training
- Story generation
- NLP research on low-resource languages
Dataset
- Language: Hindi
- Source Dataset: Simple Stories
- Content: Short, simple stories
- Format: Plain text
- Translation: English → Hindi
Translation Process
The dataset was created by translating the original Simple Stories dataset into Hindi.
The translation process aims to preserve:
- the original story structure,
- semantic meaning,
- readability,
- simple vocabulary.
Some proper names and cultural references remain close to the original dataset when appropriate.
Intended Uses
This dataset is intended for:
- Pretraining decoder-only language models
- Language modeling
- Hindi text generation
- Tokenizer development
- Academic NLP research
Limitations
- This is a translated dataset rather than stories originally written in Hindi.
- Some translated sentences may retain characteristics of the source language.
- Certain names or cultural references may not be fully localized.
Acknowledgements
This dataset is derived from the Simple Stories dataset. Please cite the original dataset alongside this translation whenever possible.
License
This dataset follows the licensing terms of the original Simple Stories dataset. Please ensure your usage complies with the original license.