---
language:
- my
license: mit
tags:
- aspect-based-sentiment-analysis
- absa
- burmese
- myanmar
- text-classification
task_categories:
- text-classification
size_categories:
- 1K<n<10K
configs:
- config_name: stage1
  data_files:
  - split: train
    path: stage1_aspect_classification.csv
  default: true
- config_name: stage2
  data_files:
  - split: train
    path: stage2_sentiment_pairs.csv
---

# Burmese Aspect-Based Sentiment Analysis (ABSA) Dataset

This dataset contains Burmese customer reviews annotated for two-stage Aspect-Based Sentiment Analysis (ABSA).

## Dataset Subsets (Configurations)

This repository contains two subsets:

1. **`stage1` (`stage1_aspect_classification.csv`)**: Multi-label aspect detection dataset containing 6 business aspects and an "All-Zero" baseline class (7,508 rows).
2. **`stage2` (`stage2_sentiment_pairs.csv`)**: Aspect-level sentiment classification pairs (Positive, Negative, Neutral).

---

## Aspect Definitions (Stage 1)
- `product_or_service_quality`: Product durability, food taste, design, cleanliness.
- `fulfillment_and_speed`: Delivery time, packaging, shipping speed.
- `price_and_value`: Value for money, discounts, price fairness.
- `digital_experience`: App/website usability, UI/UX, online payment systems.
- `customer_support`: Staff friendliness, responsiveness, helpline support.
- `variety_and_availability`: Stock availability, color/size options, menu choices.

## Sentiment Classes (Stage 2)
- `Positive`
- `Negative`
- `Neutral`

---

## How to Load in Python

```python
from datasets import load_dataset

REPO_ID = "Fixaro/burmese-aspect-based-sentiment"

# 1. Load Stage 1 (Aspect Detection)
dataset_stage1 = load_dataset(REPO_ID, "stage1")
print("Stage 1 Sample:", dataset_stage1['train'][0])

# 2. Load Stage 2 (Aspect Sentiment Analysis)
dataset_stage2 = load_dataset(REPO_ID, "stage2")
print("Stage 2 Sample:", dataset_stage2['train'][0])