CMLI-NLP/Mongolian-pretrain-dataset

Mongolian Pretraining Dataset

Dataset Information

• Language: Mongolian (Traditional Mongolian script)
• Size: ~12GB
• Format: Plain text (.txt)
• Use Case: Language model pretraining

Description

This dataset contains Mongolian text data for training language models on low-resource languages. The data uses Traditional Mongolian script and covers 45 core characters identified through frequency analysis.

Code: The Huffman transliteration framework implementation is available at https://github.com/CMLI-NLP/HuffmanTranslit

Technical Details

• Character Set: 45 core Traditional Mongolian characters
• Encoding: UTF-8
• Script: Traditional Mongolian script
• Total Samples: 933,941

Usage

Suitable for:

• Continued pretraining of multilingual language models
• Cross-lingual transfer learning research
• Low-resource language processing

Transliteration Framework Compatibility

This dataset is designed to work with the Huffman-based transliteration framework for reversible conversion between Traditional Mongolian script and Latin characters.

Citation

If you use this dataset, please cite:

@inproceedings{zhuang2025enhancing,
  title={Enhancing Cross-Lingual Transfer through Reversible Transliteration: A Huffman-Based Approach for Low-Resource Languages},
  author={Zhuang, Wenhao and Sun, Yuan and Zhao, Xiaobing},
  booktitle={Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  pages={16299--16313},
  year={2025}
}