thutasann/mm_sentiment_model_v1_data

Myanmar Sentiment Intensity Dataset v1

This dataset contains Myanmar (Burmese) text samples annotated for sentiment intensity across 5 levels. It is specifically curated for training and evaluating text classification models.

Dataset Structure

The dataset consists of two primary columns:

• text: Raw Myanmar text strings.
• label: Sentiment intensity score (Integer 0 to 4).

Label Definitions

Label	Emoji	Sentiment	Description
0	🔴	Very Negative	Strong dissatisfaction, anger, or hate
1	🟠	Negative	Dissatisfaction, boredom, or sadness
2	🟡	Neutral	Normal, factual, or indifferent
3	🟢	Positive	Satisfied, good, or peaceful
4	🔵	Very Positive	Very happy, excellent, or joyful

Data Preprocessing

To achieve optimal results with this dataset, it is highly recommended to perform Syllable Breaking before tokenization. This ensures the model processes the text using Myanmar syllable units.

Syllable Breaking Logic:

The regex implementation is based on the ye-kyaw-thu/sylbreak algorithm.

import re

def myanmar_sylbreak(line):
    pat = re.compile(r"((?<!္)[က-အ](?![်္])|[a-zA-Z0-9ဣဤဥဦဧဩဪဿ၌၍၏၀-၉၊။!-/:-@[-`{-~\s])")
    line = re.sub(r'\s+', ' ', str(line).strip())
    return pat.sub(r" \1", line).strip()

Acknowledgments & Credits

This dataset is developed by building upon existing Myanmar NLP resources and research:

• Base Dataset Source: This work utilizes data and concepts from the Myanmar Text Segmentation Dataset by Chuu Htet Naing.
• Syllable Segmentation: The syllable breaking logic is based on the sylbreak repository by Ye Kyaw Thu.