kalixlouiis/myanmar-sentiment-analysis

Dataset Description

This dataset consists of 1,665 annotated Myanmar language sentences designed for Sentiment Analysis tasks. The dataset is balanced across three sentiment categories, providing a robust foundation for training and fine-tuning machine learning models to understand Myanmar linguistic nuances, including sarcasm and compound sentence structures.

• Total size: 1,665 records
• Languages: Burmese (Myanmar)
• Task: Sentiment Classification (Positive, Negative, Neutral)

Data Statistics

Label	Count
Positive	555
Negative	555
Neutral	555
Total	1,665

---

Annotation Guideline

To ensure consistency, the dataset was annotated following these strict guidelines focusing on the Intended Meaning rather than literal translation.

1. Sentiment Categories

• Positive: Expressions of praise, satisfaction, encouragement, or happiness.
• Negative: Expressions of criticism, dissatisfaction, anger, or disappointment.
• Neutral: Fact-based information without emotional bias, or balanced statements where positive and negative sentiments offset each other.

2. Decision Rules

• Sarcasm: Sentences that are literally positive but convey a critical tone are labeled as Negative.
• Compound Sentences: In sentences containing conjunctions like "ဒါပေမဲ့" (but/however), the sentiment following the conjunction is considered the dominant intent and determines the label.
• Neutrality: When positive and negative sentiments are perfectly balanced, the sentence is labeled as Neutral.

3. Examples

Text	Label	Note
ဒီဆိုင်က ဝန်ဆောင်မှု အရမ်းကောင်းတယ်။ လာရကျိုးနပ်ပါတယ်။	Positive	Direct praise
မနေ့က ညစာမှာ မှာလိုက်တဲ့ဟင်းက အရမ်းကြာတယ်၊ ပိုက်ဆံကုန်တာပဲ အဖတ်တင်တယ်။	Negative	Direct dissatisfaction
ဒီနေ့ရုံးပိတ်ရက်ဖြစ်လို့ ရုံးခန်းတွေ အားလုံးပိတ်ထားပါတယ်။	Neutral	Factual information
"ကောင်းလိုက်တဲ့ ဝန်ဆောင်မှုပဲ" လို့ ပြောပြီးတော့ နောက်တစ်ခါ ဘယ်တော့မှ မလာတော့ဘူး။	Negative	Sarcasm detected
ပစ္စည်းကတော့ အရည်အသွေးကောင်းပါတယ်၊ ဒါပေမဲ့ ပို့ဆောင်မှု အရမ်းနှေးတယ်။	Negative	Compound sentence (conjunction focus)

---

Intended Use

This dataset is designed for researchers and developers working on:

• Building sentiment analysis models for the Myanmar language.
• Fine-tuning Large Language Models (LLMs) to better understand Burmese colloquialisms.
• Social media content monitoring and customer feedback analysis.

License

This dataset is licensed under the Creative Commons Attribution 4.0 International (CC-BY-4.0) license. It is open for everyone to use, modify, and distribute, provided that appropriate credit is given.

Citation

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

@dataset{kalixlouiis_myanmar_sentiment_analysis,
  author = {Khant Sint Heinn},
  title = {Myanmar Sentiment Analysis Dataset},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{[https://huggingface.co/datasets/kalixlouiis/myanmar-sentiment-analysis](https://huggingface.co/datasets/kalixlouiis/myanmar-sentiment-analysis)}},
  license = {CC-BY-4.0}
}

Credits

Creator: Khant Sint Heinn(Kalix Louis)