DatarrX/Myanmar-Written-Spoken-Parallel-Corpus

Myanmar Written-Spoken Parallel Corpus (MWSPC)

Dataset Description

Myanmar Written-Spoken Parallel Corpus (MWSPC) is a high-quality open-source dataset designed to bridge the gap between formal written Burmese and daily spoken Burmese. This dataset is crucial for building natural-sounding AI models that understand the linguistic nuances of the Myanmar language.

• Curated by: Khant Sint Heinn (Kalix Louis)
• Organization: DatarrX | ဒေတာ-အက်စ်
• Language: Burmese (Myanmar)
• License: Creative Commons Attribution 4.0 International (CC-BY-4.0)

What is Written vs. Spoken Burmese?

Burmese is a diglossic language, meaning there is a significant difference between the formal literary style and the informal colloquial style.

• Written Style (Literary/Formal): Used in news, textbooks, legal documents, and official speeches. It uses specific grammatical markers like "သည်" (thi), "၏" (ei), and "၍" (yway).
• Spoken Style (Colloquial/Informal): Used in daily conversations, social media, and fiction dialogue. It uses markers like "တယ်" (tal), "ရဲ့" (yae), and "ပြီး" (pyee).

AI models trained only on formal texts often sound unnatural to native speakers. MWSPC provides direct mapping between these two styles to enable "Style Transfer" and "Natural Language Understanding."

Dataset Structure

The dataset consists of 5,555 rows of parallel text pairs.

Field	Description
written_style	Formal/Literary version of the sentence (Formal Burmese)
spoken_style	Informal/Colloquial version of the sentence (Spoken Burmese)

Data Quality

Every row in this dataset has been strictly filtered to ensure 100% uniqueness. All duplicate entries have been removed, resulting in 5,555 high-quality, unique parallel text pairs.

Development History

This dataset was built by aggregating and refining:

1. kalixlouiis/myanmar-written-spoken-text-pairs (1,643 rows).
2. Additional Curated Data: 3,912 new rows added to ensure high diversity and accuracy.

Uses

Direct Use

• Style Transfer: Converting formal news or documents into casual spoken language for chatbots.
• Machine Translation: Improving the fluency of translations from English to Burmese.
• Preprocessing: Normalizing social media text into formal Burmese for downstream NLP tasks.

Dataset Creation

Curation Rationale

In the current AI era, most Myanmar datasets are scraped from news sites, which are heavily biased towards the "Written Style." This creates a lack of data for "Spoken Style" applications. MWSPC was created to empower the Burmese AI ecosystem with a data-rich foundation for next-generation innovation.

Bias, Risks, and Limitations

While the dataset covers various patterns, Burmese is a dialect-rich language. This dataset primarily focuses on the standard dialect (Yangon/Mandalay). Users should be aware that regional slang or extremely casual internet lingo may not be fully represented.

Citation

If you use this dataset in your research or project, please cite it as follows:

APA: Khant Sint Heinn, (2026). Myanmar Written-Spoken Parallel Corpus (MWSPC). DatarrX Foundation. Retrieved from [https://huggingface.co/datasets/DatarrX/Myanmar-Written-Spoken-Parallel-Corpus]

BibTeX:

@dataset{mwspc,
  author       = {Khant Sint Heinn (Kalix Louis)},
  title        = {Myanmar Written-Spoken Parallel Corpus (MWSPC)},
  year         = {2026},
  publisher    = {Hugging Face},
  organization = {DatarrX},
  url          = {https://huggingface.co/datasets/DatarrX/MWSPC}
}

---

About the Author

Khant Sint Heinn, working under the name Kalix Louis, is a Machine Learning Engineer focused on Natural Language Processing (NLP), data foundations, and open-source AI development. His work is centered on improving support for the Burmese (Myanmar) language in modern AI systems by building high-quality datasets, practical tools, and scalable infrastructure for language technology.

He is currently the Lead Developer at DatarrX, where he develops data pipelines, manages large-scale data collection workflows, and helps create open-source resources for researchers, developers, and organizations. His experience includes data engineering, web scripting, dataset curation, and building systems that support real-world machine learning applications.

Khant Sint Heinn is especially interested in advancing low-resource languages and making AI more accessible to underrepresented communities. Through his open-source contributions, he works to strengthen the Burmese (Myanmar) tech ecosystem and provide reliable building blocks for future language models, search systems, and intelligent applications.

His goal is simple: to turn limited language resources into practical opportunities through clean data, useful tools, and community-driven innovation.

Connect with the Author:
GitHub | Hugging Face | Kaggle