license: cc-by-sa-4.0
language:
- my
pretty_name: Burmese Text Corpus for Natural Language Processing
task_categories:
- text-classification
- fill-mask
- text-generation
tags:
- Myanmar
- Burmese
- NLP
size_categories:
- 1K<n<10K
Burmese Text Corpus For Natural Language Processing
ð« Choose your language: ð English Version | ð²ð² ááŒááºáá¬áá¬ážááŸááºáž
ð English Version
This dataset is a specifically curated text corpus for the Burmese language. It is intended to support Natural Language Processing (NLP) tasks, language model training, and research related to the Burmese language.
Table of Contents
- 1. About the Dataset
- 2. Dataset Structure
- 3. Data Collection
- 4. Language Specifics
- 5. Ethical Considerations & Biases
- 6. License
- 7. Acknowledgements & Citation
- 8. Issues and Contact
1. About the Dataset
The primary goal of creating this burmese-text-corpus dataset is to address the scarcity of high-quality resources available for Burmese Natural Language Processing (NLP). It aims to foster the advancement of AI models and research initiatives specifically tailored for the Burmese language.
What's included?
- This dataset contains Burmese texts collected from various websites across diverse domains. News websites are predominantly featured as their articles generally exhibit accurate spelling and an easy-to-understand writing style.
- Each data entry uniquely represents a complete and grammatically sound Burmese sentence.
- All texts within the dataset exclusively utilize Burmese characters and vocabulary; no other languages are included.
- A strong emphasis has been placed on ensuring the grammatical correctness and precise spelling of every collected sentence.
Language: Burmese (Myanmar)
Current Size
Currently, the dataset comprises a total of 3,030 (3K) data entries. This dataset is continuously being expanded, and the total data volume is expected to increase in the future.
2. Dataset Structure
The texts in this dataset are stored in JSON Lines (NDJSON) format. It includes the following fields:
text: A sentence written in the Burmese language (string type).source: The URL of the website from which the text was retrieved (string type).
Splits:
This dataset does not currently have predefined splits such as "train," "validation," or "test." All data will be available as a single "train" split.
Example Data:
{
"text": "áá¯áá¹ááá²á· áá±ááºááŒáá¯áẠáá±á«áºááœááºážáá²á·áá²á· áááááá¯ááºáᶠáá±á
áºááá¯á·á
áºáá»áœááºážá á¡áá»áá¯ážááá®ážá
á¬ááᯠáááºááá¯áá²á· áá²áá
áºááá»á¬ááœá±ááᯠá¡ááŸá
Ạáá
áá áá»á±á¬áºáá²á· áá¯áá±ááºá¡áá ááá¹áá¬áááœáŸá¬ážááŸá¬ áááºááŸá¯áá±ááŒáá«áááºá",
"source": "https://www.bbc.com/burmese/in-depth-48663133"
}
{
"text": "áá»áááºááá¯ááºáá¶ááŸá¬ ááá
áºááŸá
Ạáááá áá±á¬ááºá ááœááºáá±á«áºáá²á·áá²á· áá±áá® áá°áá¬áá¬áá®áá²á· áááºáá»á®áá¯á¶ááŒáẠááœá±ááá¯áááºáž ááá¹ááááºá
á¬áá±ááᯠá¡ááŒá
Ạáááºáá¶áá¬ážááŒáá«áááºá",
"source": "https://www.bbc.com/burmese/in-depth-48663133"
}
{
"text": "ááŒáááááºááŸá¬áá±á¬á· á¡ááŸá
Ạááá á á
á¬áá±ážááá¬á áá»áááºážá¡á±á¬áºá
áááºáá²á· ááá¹áá¯ááœá±ááᯠá
áœá²á
áœá²áááºážáááºážáááºáá°ááœá± ááŸááá±áá±ážáááᯠáá°á·áá¯ááºáá¯á¶áá²á· áá¯ááºááŒá¯ááŒá®áž ááœá±á
áá¹áá° áá¯ááºáá¬ážáá«áááºá",
"source": "https://www.bbc.com/burmese/in-depth-48663133"
}
3. Data Collection
The texts within this burmese-text-corpus dataset were collected using a combined approach of both Web Scraping and Manual curation.
Collection Process Steps:
- Web Scraping: Initially, textual data was automatically collected from various Burmese websites using web scraping techniques. This stage involved removing HTML tags and unnecessary formatting, and segmenting the content into individual sentences.
- Manual Review & Curation: The
raw dataobtained from web scraping underwent careful manual review. During this crucial step, emphasis was placed on data quality, and the following aspects were thoroughly checked:- Grammar and Spelling Accuracy: Based on Myanmar spelling guidelines, each sentence was meticulously checked and corrected for grammatical and spelling accuracy. Efforts were made to minimize errors as much as possible.
- Meaningful Completeness: Each data entry was ensured to represent a complete and meaningful Burmese sentence.
- Formatting: Each text was incorporated as a
textfield, and the URL of the originating website was assigned to thesourcefield, systematically converting the data into JSON Lines format (each line as a distinct JSON object).
Sources:
The sources from which texts were collected for this dataset are diverse:
- Primarily, a significant portion of the data was sourced from news websites (e.g., BBC Burmese, VOA Burmese).
- Additionally, educational websites, official state websites, and content published in Burmese by international organizations might also be included.
- Efforts were made to gather texts from a wide range of domains.
Through these dedicated efforts, a high-quality Burmese text dataset has been created.
Unique Characteristics and NLP Challenges of the Burmese Language
The Burmese language possesses unique characteristics that set it apart from other languages. These features can pose challenges for NLP tasks, but this dataset aims to aid in addressing them.
- Lack of Word Segmentation: Unlike English, where words are separated by spaces, Burmese texts typically do not have spaces between words. Identifying where one word begins and another ends (word segmentation) is one of the biggest challenges for Burmese NLP.
- Grammatical Structure: Burmese grammar rules can be complex, and its sentence structures may differ significantly from those of other languages.
- Limited Resources: High-quality and large-scale Burmese text datasets are not as abundant as those for many other languages.
How This Dataset Can Help:
The burmese-text-corpus dataset can contribute to overcoming the aforementioned challenges in the following ways:
- High-Quality Texts: By meticulously curating texts with accurate grammar and minimal spelling errors, this dataset can lead to more precise results when training NLP models for tasks such as
word segmentationandPart-of-Speech Tagging. - Human-like/Natural Tone: The texts in this dataset are sourced directly from websites, reflecting natural, human-like writing styles and easily understandable language usage. This characteristic will help train Language Models to better comprehend and generate more realistic Burmese language.
- Diverse Content: Including texts from news, educational, and other domains enables the training of models to achieve a broader understanding across various subject areas.
5. Ethical Considerations & Biases
When utilizing any dataset, it is crucial to be aware of potential biases and ethical considerations. Users of the burmese-text-corpus dataset are advised to keep the following points in mind:
- Source-Induced Biases:
- The majority of texts in this dataset are collected from news websites, state-affiliated websites, and international organizations' websites. Consequently, these texts may reflect the perspectives, views, and reporting styles of their respective organizations, potentially containing certain political, social, or regional biases in their content, opinions, and terminology.
- We have included the
sourceURLs, allowing users to verify the original sources and independently assess any potential biases. - The primary objective of this dataset is to facilitate the study of text structure, grammatical correctness, and Burmese vocabulary usage, as well as to train NLP models.
- Language Usage Bias:
- The texts within this dataset primarily represent Modern Burmese language writing styles. Therefore, the inclusion of regional dialects or archaic language may be limited.
- Again, the main purpose of this dataset is to focus on the structural integrity, grammatical accuracy, and lexical usage of Burmese texts for NLP model training.
- Impact on Models:
- AI models (e.g., Language Models) trained with this dataset may inherit biases present in the data. Thus, it is extremely important to interpret and evaluate the results from such models with caution.
- Sensitive Content:
- While utmost care has been taken to avoid the inclusion of sensitive and inappropriate content such as hate speech or violence within the dataset, users should be aware that political texts and political opinions may be present. We have made every effort during the dataset curation process to remove content that exhibits political bias.
The inclusion of
sourceURLs helps users to cross-reference the original content if needed.
6. License
This burmese-text-corpus dataset is released under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) License.
This means that when you use this dataset, you must give appropriate credit to the original creator (kalixlouiis / Kalix Louis). Furthermore, if you adapt, transform, or build upon the material (e.g., creating new datasets, models), you must distribute your contributions under the same CC BY-SA 4.0 license as the original.
For more details, please visit the Creative Commons Attribution-ShareAlike 4.0 International License website.
7. Acknowledgements & Citation
It was a truly joyful experience for me to create this burmese-text-corpus dataset. I sincerely hope that this dataset will contribute to the advancement of the Burmese NLP field. I welcome all users who will utilize this dataset, and I am open to any suggestions or critiques. I believe this will be beneficial for everyone.
If you use this burmese-text-corpus dataset in your research, projects, or any other work, please cite the original creator using the following format:
@misc{kalixlouiis-burmese-text-corpus_2025,
author = {Kalix Louis},
title = {Burmese Text Corpus for Natural Language Processing},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/kalixlouiis/burmese-text-corpus}
}
8. Issues and Contact
If you have any questions or encounter any issues (e.g., data errors, biases) related to this dataset, please feel free to report them via the "Discussions" tab on the Hugging Face Repository.
Your suggestions and reports are highly welcome, as they will help us in improving the quality of this dataset.
ð²ð² ááŒááºáá¬áá¬ážááŸááºáž
áá® dataset áá±ážááá±á¬á· ááŒááºáá¬áá¬áá¬á áá¬ážá¡ááœáẠá¡áá°ážáááºááœááºááŒá®áž á á¯á ááºážáá¬ážáá²á· á á¬áá¬ážáá±á«ááºážáá»á¯áẠ(Text Corpus) áá áºáᯠááŒá áºáá«áááºá Natural Language Processing (NLP) áá¯ááºáááºážá á¥áºá áá¬áá¬á áá¬ážáá±á¬áºáááºáá»á¬áž áá±á·áá»áá·áºáá±ážááŒááºážááŸáá·áº ááŒááºáá¬áá¬áá¬á áá¬ážááá¯ááºáᬠáá±á·áá¬ááŸá¯áá»á¬ážá¡ááœáẠá¡áá±á¬ááºá¡áá°ááŒá¯ááẠáááºááœááºáá«áááºá
áá¬áááá¬
- áá Dataset á¡ááŒá±á¬ááºáž
- áá Dataset áááºáá±á¬ááºáá¯á¶
- áá Data ááœá±ááᯠáááºááá¯á á¯áá±á¬ááºážáá²á·áá²á
- áá áá¬áá¬á áá¬ážááŸáá·áº áááºááá¯ááºáᬠá¡áá»ááºá¡áááº
- á á áá»áá·áºáááºááá¯ááºáᬠá ááºážá á¬ážááŸá¯áá»á¬áž ááŸáá·áº áááºááá¯ááºááŸá¯áá»á¬áž
- áá ááá¯ááºá ááº
- áá áá»á±ážáá°ážáááºááœáŸá¬ ááŸáá·áº ááá¯ážáá¬ážáá»ááºáá»á¬áž
- áá ááŒá¿áá¬áá»á¬áž ááŸáá·áº áááºááœááºáááº
áá Dataset á¡ááŒá±á¬ááºáž
áá® burmese-text-corpus dataset ááᯠáááºáá®ážáááŒááºážáá²á· á¡ááááááºááœááºáá»ááºááá±á¬á· ááŒááºáá¬áá¬áá¬á
áá¬áž Natural Language Processing (NLP) áááºáááºááŸá¬ ááá¯á¡ááºáá±áá²á· á¡áááºá¡ááœá±ážááŒáá·áº á¡áááºážá¡ááŒá
áºááœá±ááᯠááŒáá·áºáááºážáá±ážááá¯á·áá² ááŒá
áºáá«áááºá ááŒááºáá¬áá¬áá¬á
áá¬ážáá²á·áááºáááºáá²á· AI áá±á¬áºáááºááœá±áá²á· áá¯áá±áááá¯ááºáááºážááœá± ááá¯ááá¯ááá¯ážáááºáá¬á
á±ááá¯á·á¡ááœáẠáááºááœááºáá«áááºá
áá¬ááœá±áá«áááºáá²á
- áá® dataset áá²ááŸá¬áá±á¬á· á¡áá»áá¯ážáá»áá¯ážáá±á¬ Website ááœá±á áááºáááºá¡á á¯á¶á¡áááºááá± ááá°áá¬ážáá²á· ááŒááºáá¬á á¬áá¬ážááœá±ááᯠá á¯á ááºážáá±ážáá¬ážáá«áááºá á¡áá»á¬ážá¡á¬ážááŒáá·áº ááááºáž Website ááœá±á á á¬áá¬ážááœá±ááᯠá¡ááááá¬ážááŒá®áž ááá·áºááœááºážáá¬ážáá«áááºá áá¬ááŒá±á¬áá·áºáá²ááá¯áá±á¬á· ááááºážá á¬áá¬ážááœá±áᬠá á¬áá¯á¶ážáá±á«ááºážáááºáá¯á¶ááŸááºáááºááŒá®áž áá¬ážáááºáááœááºáá°áá²á· áá±ážáá¬ážááŸá¯ááœá± ááŸááá±ááá¯á·áá«á
- Data áá áºáá¯áá»ááºážá á®áᬠá¡áááá¹áá¬ááºááŒáá·áºá á¯á¶áá²á· ááŒááºáá¬á á¬ááŒá±á¬ááºážáá áºááŒá±á¬ááºážááᯠááá¯ááºá á¬ážááŒá¯áá«áááºá
- Dataset áá²ááŸá¬áá«áá²á· á á¬áá¬ážááœá±á¡á¬ážáá¯á¶ážáᬠááŒááºáá¬á¡áá¹ááá¬ááœá±áá²á· ááŒááºáá¬áá±á«áá¬áááœá±áááºáááºáá² ááŒá áºááŒá®ážá á¡ááŒá¬áž áá¬áá¬á áá¬ážááœá± áá¯á¶ážááá«áááºááŒááºáž áááŸááá«áá°ážá
- á á¯á ááºážáá¬ážáá²á· á á¬ááŒá±á¬ááºážááá¯ááºážáᬠááá¹áá«ááŸááºáááºááá¯á·áá²á· á á¬áá¯á¶ážáá±á«ááºážáááºáá¯á¶ááŸááºáááºááá¯á·ááᯠáŠážá á¬ážáá±ážáá¬ážáá«áááºá
áá¬áá¬á áá¬áž: ááŒááºáᬠ(ááá¬)
áááºááŸá á¡ááœááºá¡á á¬áž
áááºááŸáá¡áá»áááºááŸá¬áá±á¬á· Data á á¯á á¯áá±á«ááºáž 3,030 (3K) ááᯠá á¯á ááºážáá¬ážááŒá®áž ááŒá áºáá«áááºá Dataset ááᯠáááºáááºááŒáá·áºáááºážáá±áá²ááŒá áºááá¯á· áá±á¬ááºááá¯ááºážááŸá¬ Data ááá¬á ááá¯ááá¯ááá¯ážáá¬ááŸá¬áá«á
áá Dataset áááºáá±á¬ááºáá¯á¶
áá® dataset ááŸá¬ á á¬áá¬áž (text) ááœá±ááᯠJSON Lines (NDJSON) format áá²á· ááááºážáááºážáá¬ážáá«áááºá á¡á±á¬ááºáá±á¬áºááŒáá« fields ááœá± áá«áááºáá«áááºá
text: ááŒááºáá¬áá¬áá¬á áá¬ážááŒáá·áº áá±ážáá¬ážáá¬ážáá±á¬ á á¬ááŒá±á¬ááºáž (string á¡áá»áá¯ážá¡á á¬áž)ásource: á¡ááá¯áá«á á¬áá¬ážááᯠááá°áá²á·áá±á¬ website á URL (string á¡áá»áá¯ážá¡á á¬áž)á
Splits:
áá® dataset ááŸá¬áá±á¬á· ááŒáá¯áááºáááºááŸááºáá¬ážáá²á· "train", "validation", "test" á áááŒáá·áº á¡ááá¯ááºážááœá± ááœá²ááŒá¬ážáá¬ážááŒááºáž áááŸááá±ážáá«áá°ážá Data á¡á¬ážáá¯á¶ážááᯠáá áºáá±á«ááºážáááºáž "default" split á¡áá±áá²á· áááŸááá«áááá·áºáááºá
á¥ááᬠData:
{
"text": "áá¯áá¹ááá²á· áá±ááºááŒáá¯áẠáá±á«áºááœááºážáá²á·áá²á· áááááá¯ááºáᶠáá±á
áºááá¯á·á
áºáá»áœááºážá á¡áá»áá¯ážááá®ážá
á¬ááᯠáááºááá¯áá²á· áá²áá
áºááá»á¬ááœá±ááᯠá¡ááŸá
Ạáá
áá áá»á±á¬áºáá²á· áá¯áá±ááºá¡áá ááá¹áá¬áááœáŸá¬ážááŸá¬ áááºááŸá¯áá±ááŒáá«áááºá",
"source": "https://www.bbc.com/burmese/in-depth-48663133"
}
{
"text": "áá»áááºááá¯ááºáá¶ááŸá¬ ááá
áºááŸá
Ạáááá áá±á¬ááºá ááœááºáá±á«áºáá²á·áá²á· áá±áá® áá°áá¬áá¬áá®áá²á· áááºáá»á®áá¯á¶ááŒáẠááœá±ááá¯áááºáž ááá¹ááááºá
á¬áá±ááᯠá¡ááŒá
Ạáááºáá¶áá¬ážááŒáá«áááºá",
"source": "https://www.bbc.com/burmese/in-depth-48663133"
}
{
"text": "ááŒáááááºááŸá¬áá±á¬á· á¡ááŸá
Ạááá á á
á¬áá±ážááá¬á áá»áááºážá¡á±á¬áºá
áááºáá²á· ááá¹áá¯ááœá±ááᯠá
áœá²á
áœá²áááºážáááºážáááºáá°ááœá± ááŸááá±áá±ážáááᯠáá°á·áá¯ááºáá¯á¶áá²á· áá¯ááºááŒá¯ááŒá®áž ááœá±á
áá¹áá° áá¯ááºáá¬ážáá«áááºá",
"source": "https://www.bbc.com/burmese/in-depth-48663133"
}
áá Data ááœá±ááᯠáááºááá¯á á¯áá±á¬ááºážáá²á·áá²á
áá® burmese-text-corpus dataset áá²á á
á¬áá¬ážááœá±ááᯠWeb Scraping áááºážáááºážáá²á·áá±á¬á áá°ááá¯ááºááá¯áẠ(Manual) ááŸá¬ááœá±áá±á¬ááºáá°áá¬áá²á·áá±á¬ áá±á«ááºážá
ááºááŒá®áž á
á¯áá±á¬ááºážáá²á·áá«áááºá
á
á¯áá±á¬ááºážáá¯á¶ á¡ááá·áºááá·áº:
á. Web Scraping: ááááŠážá
áœá¬á á¡áá»áá¯ážáá»áá¯ážáá±á¬ ááŒááºáá¬áááºááºááá¯ááºááœá±ááá± á
á¬áá¬ážá¡áá»ááºá¡áááºááœá±ááᯠWeb Scraping áááºážáááºážáá²á· á¡ááá¯á¡áá»á±á¬áẠá
á¯áá±á¬ááºážáá²á·áá«áááºá áá®á¡ááá·áºááŸá¬ HTML tags ááœá±á áááá¯á¡ááºáá²á· áá±á¬áºáááºááœá± áááºááŸá¬ážááŒá®áž á
á¬ááŒá±á¬ááºážááœá±ááᯠááœá²áá¯ááºáá¬áá»áá¯ážááœá± áá«áááºáá«áááºá
á. Manual Review & Curation: Web Scraping ááá± áááŸááá¬áá²á· raw data ááœá±ááᯠáá°ááá¯ááºááá¯áẠááá¯áá
áá¯áẠááŒááºáááºá
á
áºáá±ážáá²á·áá«áááºá áá®á¡ááá·áºááŸá¬ data áá²á· á¡áááºá¡ááœá±ážááᯠá¡ááá áŠážá
á¬ážáá±ážááŒá®áž á¡á±á¬ááºáá«á¡áá»ááºááœá±ááᯠáá±áá»á¬á
á
áºáá±ážáá²á·áá«áááºá
* ááá¹áá«ááŸáá·áº á
á¬áá¯á¶ážáá±á«ááºážáááºáá¯á¶ ááŸááºáááºááŒááºáž: á
á¬áá¯á¶ážáá±á«ááºážáááºáá¯á¶áá»ááºážááᯠá¡ááŒá±áá¶ááŒá®áž á
á¬ááŒá±á¬ááºážááá¯ááºážáá²á· ááá¹áá«áá²á· á
á¬áá¯á¶ážáá±á«ááºážáááºáá¯á¶ááœá± ááŸááºáááºááŸá¯ááŸááááŸáááᯠáá±áá»á¬á
á
áºáá±ážááŒááºáááºáá²á·áá«áááºá á¡ááŸá¬ážá¡ááœááºážááœá±ááᯠá¡áááºááá¯ááºáá¯á¶áž áá»áŸá±á¬á·áá»áá¬ážáá«áááºá
* á¡áááá¹áá¬ááºááŒáá·áºá
á¯á¶ááŒááºáž: Data áá
áºáá¯áá»ááºážá
á®áᬠá¡áááá¹áá¬ááºááŒáá·áºá
á¯á¶áá²á· ááŒááºáá¬á
á¬ááŒá±á¬ááºáž ááŒá
áºáá±ááá¯á· áá±áá»á¬á
á±áá²á·áá«áááºá
* Format ááŒá¯áá¯ááºááŒááºáž: á
á¬áá¬ážáá
áºáá¯á
á®ááᯠtext field á¡ááŒá
Ạááá·áºááœááºážááŒá®ážá ááá°áá²á·áá²á· website URL ááá¯áá±á¬á· source field á¡áá±áá²á· JSON Lines format (áá
áºááŒá±á¬ááºážáá»ááºážá
á® JSON object áá¯á¶á
á¶) ááŒá
áºá¡á±á¬áẠá
áá
áºááá» ááŒá¯áá¯ááºáá²á·áá«áááºá
áááºážááŒá áºáá»á¬áž (Sources):
áá® dataset á¡ááœáẠá á¬áá¬ážááœá±ááᯠááá°áá²á·áá²á· á¡áááºážá¡ááŒá áºááœá±ááá±á¬á· á¡áá»áá¯ážáá»áá¯ážááŸááá«áááºá
- á¡áááá¡á¬ážááŒáá·áº ááááºážáááºááºááá¯ááºáá»á¬áž (á¥ááá¬: BBC Burmese, VOA Burmese á áááŒáá·áº) ááá± ááá°áᬠáá»á¬ážáá«áááºá
- áá«á·á¡ááŒáẠááá¬áá±ážááá¯ááºáᬠáááºááºááá¯ááºáá»á¬ážá ááá¯ááºáá¶áá±á¬áºáá²á· áááºááºááá¯ááºááœá±áá²á· ááá¯ááºáá¶ááᬠá¡ááœá²á·á¡á ááºážááœá±áá²á· ááŒááºáá¬áá¬áá¬áá²á· áá¯ááºááŒááºáá¬ážáá²á· á¡áá»ááºá¡áááºááœá±áááºáž áá«áááºááá¯ááºáá«áááºá
- áááºáááºá¡á á¯á¶á¡áááºá á á¬áá¬ážááœá±ááᯠáááŸáááá¯á· ááŒáá¯ážá á¬ážáá¬ážáá«áááºá
áá®ááá¯áá»áá¯áž ááŒáá¯ážá á¬ážá¡á¬ážáá¯ááºááŸá¯ááœá±ááá± á¡áááºá¡ááœá±ážáá±á¬ááºážááœááºáá²á· ááŒááºáá¬á á¬áá¬áž dataset áá áºáá¯ááŒá áºáá¬á¡á±á¬áẠáááºáá®ážáá¬ážáá«áááºá
áá áá¬áá¬á áá¬ážááŸáá·áº áááºááá¯ááºáᬠá¡áá»ááºá¡áááº
ááŒááºáá¬áá¬áá¬á áá¬ážáá²á· áá°ážááŒá¬ážáá»ááºáá»á¬áž ááŸáá·áº NLP á áááºáá±á«áºááŸá¯áá»á¬áž
ááŒááºáá¬áá¬áá¬á áá¬ážáᬠá¡ááŒá¬ážáá¬áá¬á áá¬ážááœá±áá²á· ááá°áá² áá°ážááŒá¬ážáá²á· áááá±áááá¹ááá¬ááœá± ááŸááá«áááºá áá®á¡áá»ááºááœá±á NLP áá¯ááºáááºážááœá±á¡ááœáẠá áááºáá±á«áºááŸá¯ááœá±áááºáž ááŒá áºááá¯ááºáááá¯á áá® dataset ááá± á¡ááŒá±ááŸá¬ááá¯á· á¡áá±á¬ááºá¡áá°ááŒá áºá á±ááá¯ááºáá«áááºá
- á áá¬ážáá¯á¶ážááœá²ááŒá¬ážááŒááºáž (Word Segmentation) áááŸáááŒááºáž: á¡ááºá¹áááááºá á¬ááᯠá áá¬ážáá¯á¶ážááœá±ááᯠspace áá²á· ááœá²ááŒá¬ážáá¬ážáá¬áá»áá¯áž ááŒááºáá¬á á¬ááŸá¬ áááŸááá«áá°ážá á á¬áá¯á¶ážááœá±á áááºáá±áááºáá²á·á¡ááœáẠáááºáá±áá¬ááá± á áá¬ážáá¯á¶ážáá áºáá¯á¶áž á ááŒá®áž áááºáá±áá¬ááŸá¬ áá¯á¶ážáááºááá¯áá¬ááᯠááœá²ááŒá¬ážááᬠNLP á¡ááœáẠá¡ááŒá®ážáá¬ážáá¯á¶áž á áááºáá±á«áºááŸá¯ááœá±áá²á áá áºáá¯áá«á
- ááá¹áá«áááºáá±á¬ááºáá¯á¶ (Grammatical Structure): ááŒááºáá¬áá¬áá¬á áá¬ážááŸá¬ áá¯á¶áá±ááá¹áá«á ááºážáá»ááºážááœá± ááŸá¯ááºááœá±ážááá¯ááºááŒá®ážá áá«áá»áááºáá±á¬ááºáá¯á¶ááááºáž á¡ááŒá¬ážáá¬áá¬á áá¬ážááœá±áá²á· ááá°áá®áá²á· áá¯á¶á á¶áá»áá¯ážááœá± ááŸáááá¯ááºáá«áááºá
- Resource áááºážáá«ážááŒááºáž: á¡áááºá¡ááœá±ážááŒáá·áºááŒá®áž ááŒá®ážáá¬ážáá²á· ááŒááºáá¬á á¬áá¬áž dataset ááœá±áᬠá¡ááŒá¬ážáá¬áá¬á áá¬ážááœá±áá±á¬áẠááá±á«áá»á¬ážáá±ážáá«áá°ážá
áá® dataset á áááºááá¯áá°áá®ááá¯ááºááá²á
burmese-text-corpus dataset áᬠá¡áááºáá±á¬áºááŒáá« á
áááºáá±á«áºááŸá¯ááœá±ááᯠááŒá±ááŸááºážááá¯á·á¡ááœáẠá¡á±á¬ááºáá«á¡ááá¯ááºáž á¡áá±á¬ááºá¡áá°ááŒá¯ááá¯ááºáá«áááºá
- á¡áááºá¡ááœá±ážááŒáá·áº á
á¬áá¬ážáá»á¬áž: ááá¹áá«ááŸááºáááºááŒá®áž á
á¬áá¯á¶ážáá±á«ááºážáááºáá¯á¶ á¡ááŸá¬ážá¡ááœááºáž á¡áááºážáá¯á¶ážááŒá
áºá¡á±á¬áẠááá¯áá
áá¯áẠááŒááºáááºáá¬ážáá¬ááŒá±á¬áá·áº
word segmentation(á áá¬ážáá¯á¶ážááœá²ááŒá¬ážááŒááºáž)áPart-of-Speech Tagging(á áá¬ážáá¯á¶ážá¡áá»áá¯ážá¡á á¬ážááœá²ááŒá¬ážááŒááºáž) á áá²á· NLP áá±á¬áºáááºááœá± áá±á·áá»áá·áºáá¬ááŸá¬ áááá»ááŸá¯ááŒáá·áºáá¬ážáá²á· ááááºááœá± áááŸáá á±ááá¯ááºáá«áááºá - ááá¬ááá»áá±á¬ áá°á·á áá¬ážáá¯á¶ážáá»á¬áž (Human-like/Natural Tone): dataset áá²á á á¬áá¬ážááœá±áᬠwebsite ááœá±ááá± ááá¯ááºááá¯ááºááá°áá¬ážáá²á· áá°áá¬ážááœá±áá²á· áá±ážáá¬ážáá¯á¶á¡ááá¯ááºáž ááá¬ááá»ááŒá®ážá áááºááŸá¯áá°ááœá± áá¬ážáááºááœááºáá²á· á áá¬ážáá¯á¶áž á¡áá¯á¶ážá¡ááŸá¯ááºážááœá± áá«áááºáá«áááºá áá«ááŒá±á¬áá·áº áá¬áá¬á áá¬ážáá±á¬áºáááºááœá± (Language Models) ááᯠáá±á·áá»áá·áºáá²á·á¡áá« ááá¯ááá¯áááºááœá±á·áá»áá²á· áá¬áá¬á áá¬ážááᯠáá¬ážáááºá áá¯ááºáá¯ááºááá¯ááºááá¯á· áá°áá®áá±ážáá«áááá·áºáááºá
- á¡áá»áá¯ážáá»áá¯ážáá±á¬ á¡ááŒá±á¬ááºážá¡áá¬áá»á¬áž: ááááºážá ááá¬áá±ážáá²á· á¡ááŒá¬ážáááºáááºáá»á¬ážá០á á¬áá¬ážáá»á¬áž áá«áááºáá²á·á¡ááœáẠáá±á¬áºáááºááœá±ááᯠááá¯ááá¯áá»ááºááŒáá·áºáá²á· á¡ááŒá±á¬ááºážá¡áá¬áááºáááºááœá±ááŸá¬ áá¬ážáááºááá¯ááºá áœááºážááŸáá¡á±á¬áẠáá±á·áá»áá·áºáá±ážááá¯ááºáá«áááºá
á á áá»áá·áºáááºááá¯ááºáᬠá ááºážá á¬ážááŸá¯áá»á¬áž ááŸáá·áº áááºááá¯ááºááŸá¯áá»á¬áž
dataset áá
áºáá¯ááᯠá¡áá¯á¶ážááŒá¯áá²á·á¡áá« ááŒá
áºááá¯ááºáá²á· áááºááá¯ááºááŸá¯ááœá± (biases) áá²á· áá»áá·áºáááºááá¯ááºáᬠá
ááºážá
á¬ážááŸá¯ááœá±ááᯠááááŸááá¬ážááá¯á·á á¡ááœááºá¡áá±ážááŒá®ážáá«áááºá burmese-text-corpus dataset ááᯠá¡áá¯á¶ážááŒá¯áá°ááœá±á¡áá±áá²á· á¡á±á¬ááºáá«á¡áá»ááºááœá±ááᯠáááááŒá¯á
á±ááá¯áá«áááºá
- áááºážááŒá
áºá០ááŒá
áºáá±á«áºáá¬áá±á¬ áááºááá¯ááºááŸá¯áá»á¬áž (Source-Induced Biases):
- áá® dataset áá²ááŸá¬áá«áááºáá²á· á á¬áá¬ážá¡áá»á¬ážá á¯ááᯠááááºážáááºááºááá¯ááºááœá±á ááá¯ááºáá¶áá±á¬áºáá²á· áááºááºááá¯ááºááœá±áá²á· ááá¯ááºáá¶ááá¬á¡ááœá²á·á¡á ááºážááœá±áá²á· áááºááºááá¯ááºááœá±ááá± á á¯áá±á¬ááºážáá¬ážáᬠááŒá áºáá«áááºá áá«ááŒá±á¬áá·áº áá®á á¬áá¬ážááœá±áᬠáááºááá¯ááºáᬠá¡ááœá²á·á¡á ááºážááœá±áá²á· áá¶áá°áá»ááºáá»á¬ážá á¡ááŒááºáá»á¬ážá ááááºážáááºááŒáá¯á¶áá»á¬ážá¡ááá¯ááºážáᬠááŒá áºáá±ááá¯ááºááŒá®ážá á¡ááŒá±á¬ááºážá¡áá¬á á¡ááŒááºáá²á· á¡áá¯á¶ážá¡ááŸá¯ááºážááœá±ááŸá¬ ááá¯ááºáá¶áá±ážá áá°ááŸá¯áá±ážá áá±áááá¯ááºáᬠáááºááá¯ááºááŸá¯á¡áá»áá¯á· áá«áááºáá¬ááá¯ááºáá«áááºá
- áá»áœááºáá±á¬áºááá¯á·á¡áá±áá²á·
sourceURL ááœá±ááᯠááá·áºááœááºážáá±ážáá¬ážáá¬ááŒá±á¬áá·áº á¡áá¯á¶ážááŒá¯áá°áá»á¬ážá¡áá±ááŒáá·áº ááá¯á¡ááºáá«á áááºážááŒá áºááᯠááŒááºáááºá á áºáá±ážááá¯ááºááŒá®áž áááºááá¯ááºááŸá¯ ááŸáá áááŸáááᯠááá¯ááºááá¯áẠáá¯á¶ážááŒááºááá¯ááºáá«áááºá - áá® dataset áá²á· á¡ááááááºááœááºáá»ááºááá±á¬á· á á¬áá¬ážááœá±áá²á· áááºáá±á¬ááºáá¯á¶á ááá¹áá«ááŸááºáááºááŸá¯áá²á· ááŒááºáá¬áá±á«áá¬á á¡áá¯á¶ážá¡ááŸá¯ááºážááœá±ááᯠáá±á·áá¬ááá¯á·áá²á· NLP áá±á¬áºáááºááœá± áá±á·áá»áá·áºááá¯á·áᬠááŒá áºáá«áááºá
- áá¬áá¬á
áá¬áž á¡áá¯á¶ážááŒá¯ááŸá¯ áááºááá¯ááºááŸá¯ (Language Usage Bias):
- áá® dataset áá²á á á¬áá¬ážááœá±áᬠáá±ááºááŒáá¯áẠ(Modern) ááŒááºáá¬áá¬áá¬á áá¬áž áá²á· áá±ážáá¬ážáá¯á¶ááᯠá¡ááá ááá¯ááºá á¬ážááŒá¯áá«áááºá áá±ááá¹ááá áá¬áž (dialects) áá»á¬áž ááá¯á·ááá¯áẠááŸá±ážáá±á¬ááºážá á¬áá± (archaic language) áá»á¬áž áá«áááºááŸá¯ áááºážáá«ážááá¯ááºáá«áááºá
- áá® dataset áá²á· á¡ááááááºááœááºáá»ááºááá±á¬á· á á¬áá¬ážááœá±áá²á· áááºáá±á¬ááºáá¯á¶á ááá¹áá«ááŸááºáááºááŸá¯áá²á· ááŒááºáá¬áá±á«áá¬á á¡áá¯á¶ážá¡ááŸá¯ááºážááœá±ááᯠáá±á·áá¬ááá¯á·áá²á· NLP áá±á¬áºáááºááœá± áá±á·áá»áá·áºááá¯á·áᬠááŒá áºáá«áááºá
- áá±á¬áºáááºáá»á¬áž á¡áá±á«áº áááºáá±á¬ááºááŸá¯ (Impact on Models):
- áá® dataset áá²á· áá±á·áá»áá·áºáá¬ážáá²á· AI áá±á¬áºáááºááœá± (á¥ááá¬: áá¬áá¬á áá¬ážáá±á¬áºáááºáá»á¬áž) ááŸá¬ dataset áá²á áá«áááºááá¯ááºáá²á· áááºááá¯ááºááŸá¯ááœá± áá«ááœá¬ážááá¯ááºáá«áááºá áá«ááŒá±á¬áá·áº á¡ááá¯áá«áá±á¬áºáááºááœá±ááá± áááŸááá¬áá²á· ááááºááœá±ááᯠá¡áá¯á¶ážááŒá¯áá²á·á¡áá« ááááá¬ážááŒá®áž á¡áááá¹áá¬ááºááœáá·áºááá¯á á¡áá²ááŒááºááá¯á· á¡áá°ážá¡áá±ážááŒá®ážáá«áááºá
- ááááá¯ááºááœááºáá±á¬ á¡ááŒá±á¬ááºážá¡áá¬áá»á¬áž (Sensitive Content):
- dataset áá²ááŸá¬ á¡áá¯ááºážá áá¬áž (hate speech)á á¡ááŒááºážáááºááŸá¯ (violence) á áá²á· ááááá¯ááºááœááºááŒá®áž áááá·áºáá»á±á¬áºáá²á· á¡ááŒá±á¬ááºážá¡áá¬ááœá± áá«áááºáá¬ááá¯ááºááŒá±ááᯠá¡áááºááá¯ááºáá¯á¶áž ááŸá±á¬ááºááŸá¬ážáá¬ážáá«áááºá ááá¯á·áá±á¬áºáááºážá ááá¯ááºáá¶áá±ážááá¯ááºáᬠá á¬áá¬ážáá»á¬ážá ááá¯ááºáá¶áá±ážá¡ááŒááºáá»á¬áž áá«áááºááá¯ááºáá¬ááŒá±á¬áá·áº á¡áá¯á¶ážááŒá¯áá°áá»á¬ážá¡áá±ááŒáá·áº ááá¯á¡ááŒá±á¬ááºážá¡áá¬áá»á¬ážááᯠáááºááá¯ááºáááá¯ááºááŒá±á¬ááºáž áááááŒá¯á á±ááá¯áá«áááºá áá»áœááºáá±á¬áºááá¯á·á¡áá±ááŒáá·áº dataset ááᯠá á áºáá±ážá ááºááœáẠááá¯ááºáá¶áá±ážá¡ááŒááºáááºááá¯ááºááŸá¯áááŸáá á±ááẠá¡áááºááá¯ááºáá¯á¶áž ááá¯ááŒá¯áááºááŸá¬ážáá¬ážáá«áááºá
sourceURL ááœá± áá«áááºáá¬á á¡áá¯á¶ážááŒá¯áá°ááœá±á¡áá±áá²á· áá°áááºážá¡ááŒá±á¬ááºážá¡áá¬ááᯠááŒááºáááºá á áºáá±ážááá¯ááºááá¯á· á¡áá±á¬ááºá¡áá°ááŒá¯áá«áááºá
áá ááá¯ááºá ááº
áá® burmese-text-corpus dataset ááᯠCreative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) ááá¯ááºá
ááºá¡á±á¬ááºááŸá¬ áá¯ááºáá±ááŒáá·áºáá»á®áá¬ážáá«áááºá
áá«á áá¬ááá¯ááá¯ááá¯ááá²ááá¯áá±á¬á·á áá® dataset ááᯠá¡áá¯á¶ážááŒá¯áá²á·á¡áá« áá°áááºážáááºáá®ážáá° kalixlouiis(Kalix Louis) ááᯠcredit (ááá¯ážáá¬áž) áá±ážááá«áááºá áá«á·á¡ááŒááºá áá® dataset áá²á data ááœá±ááᯠáá°ááŒá®áž ááŒá¯ááŒááºááœááºážáá¶áá¬á á¡ááŒá±áá¶ááŒá®áž áááºáá®ážáá¬ážáá²á· á¡áá¬ááœá± (á¥ááá¬- dataset á¡áá
áºá áá±á¬áºáááºáá»á¬áž) ááᯠááŒáá·áºáá±áááºááá¯áááºáááºáž áá°áááºáž CC BY-SA 4.0 ááá¯ááºá
ááºá¡á±á¬ááºááŸá¬áá² ááŒááºáááºáá»áŸáá±ááá«áááºá
á¡áá±ážá áááºááá¯áá±á¬á· Creative Commons Attribution-ShareAlike 4.0 International License áááºááºááá¯ááºááŸá¬ áááºáá±á¬ááºááŒáá·áºááŸá¯ááá¯ááºáá«áááºá
áá áá»á±ážáá°ážáááºááœáŸá¬ ááŸáá·áº ááá¯ážáá¬ážáá»ááºáá»á¬áž
áá® burmese-text-corpus dataset áá±ážááᯠáááºáá®ážáá²á·ááᬠáá»áœááºáá±á¬á·áºá¡ááœáẠáááá·áºááᯠáá»á±á¬áºááœáŸááºá
áá¬áá«á
áá® dataset ááá± ááŒááºáᬠNLP áááºáááºááᯠááá¯ážáááºá¡á±á¬áẠáá°áá®ááá¯ááºáááºááá¯á· áá»áŸá±á¬áºááá·áºáá«áááºá
áá® dataset ááᯠá¡áá¯á¶ážááŒá¯ááá·áºáá°ááœá± á¡á¬ážáá¯á¶ážááá¯áááºáž ááŒáá¯ááá¯áá«áááºá
á¡ááŒá¶ááŒá¯áá»ááºááœá±á áá±áááºááŸá¯ááœá±ááá¯áááºáž ááŒáá¯ááá¯áá«áááºá á¡á¬ážáá¯á¶áž á¡áá»áá¯ážááŸááááºááá¯á· áá¯á¶ááŒááºáá«áááºá
á¡áááºá áááºááẠဠburmese-text-corpus dataset ááᯠáááºá áá¯áá±áááá¯ááºáááºážá ááá±á¬áá»áẠááá¯á·ááá¯áẠá¡ááŒá¬ážáá±á¬ áá¯ááºáááºážáá»á¬ážááœáẠá¡áá¯á¶ážááŒá¯áá²á·áá«á áá°áááºážáááºáá®ážáá°á¡á¬áž á¡á±á¬ááºáá«áá¯á¶á
á¶á¡ááá¯ááºáž ááá¯ážáá¬ážáá±á¬áºááŒáá±ážáá«ááẠáá±áá¹áá¬áááºáá¶á¡ááºáá«áááºá
@misc{kalixlouiis-burmese-text-corpus_2025,
author = {Kalix Louis},
title = {Burmese Text Corpus for Natural Language Processing},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/kalixlouiis/burmese-text-corpus}
}
(ááŸááºáá»ááº- BibTeX format ááᯠá¡áá¯á¶ážáááŒá¯áá«á "Kalix Louis (2025). Burmese Text Corpus for Natural Language Processing. Hugging Face. Retrieved from https://huggingface.co/datasets/kalixlouiis/burmese-text-corpus" áᯠááá¯ážáá¬ážáá±á¬áºááŒááá¯ááºáá«áááºá)
áá ááŒá¿áá¬áá»á¬áž ááŸáá·áº áááºááœááºáááº
áá® dataset áá²á·áááºáááºááŒá®áž áá±ážááŒááºážá áá¬áá»á¬ážá ááŒá¿áá¬áá»á¬áž (á¥ááá¬: data á¡ááŸá¬ážá¡ááœááºážáá»á¬ážá áááºááá¯ááºááŸá¯áá»á¬áž) ááœá±á·ááŸááá²á·áá«á Hugging Face Repository áá²á· "Discussions" tab ááá± ááŒá±á¬ááŒá¬ážáá±ážááá¯ááºáá«áááºá
ááá·áºáá²á· á¡ááŒá¶ááŒá¯áá»ááºááœá±áá²á· ááŒá±á¬ááŒá¬ážáá»ááºááœá±ááᯠááŒáá¯ááá¯áá«áááºá áá«á dataset áá²á· á¡áááºá¡ááœá±ážááᯠááŒáŸáá·áºáááºáá¬ááŸá¬ á¡áá±á¬ááºá¡áá°ááŒá áºá á±ááŸá¬áá«á