Atha Text Dataset: Indonesian Sentiment Classification for NLP Experiments

Atha Text Dataset is a sentiment classification resource for the Indonesian language, containing three sentiment classes. The dataset is authored by Bangkah and was last updated on April 13, 2026. Its intended purpose is for learning NLP pipelines and establishing experimental baselines, not for production benchmarking.

Use Cases

• Train a sentiment classifier based on the three labeled categories (negative, neutral, positive) mentioned in the description.
• Establish baseline performance metrics for Indonesian NLP experiments based on the provided train/validation splits.
• Learn and prototype text classification pipelines using a structured dataset with text and label columns.
• Compare model performance on a non-English language sentiment task based on the Indonesian text data.

Strengths

• Provides explicit three-class sentiment labels (negative, neutral, positive) for each text entry.
• Includes both numeric and text label formats (label and label_text columns) for flexibility.
• Contains predefined train and validation splits for immediate experimental use.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.
• The author states it is for learning and baselines, not production benchmarking, suggesting potential scale or quality constraints.

Provenance

Source

huggingface

Collection Method

Likely collected from Indonesian language reviews or social media, but the specific gathering method is not detailed.

Time Range

null

Freshness

Last updated 2026-04-13 04:47:15; freshness should be verified.

Geography

null