KurdABSA: Kurdish aspect-based sentiment analysis dataset curation using few-shot learning.

Area of Science:

• Natural Language Processing (NLP)
• Computational Linguistics
• Sentiment Analysis

Background:

• Aspect-Based Sentiment Analysis (ABSA) enhances traditional sentiment analysis by identifying sentiment towards specific aspects.
• Research in ABSA is extensive but lacks resources for low-resource languages like Kurdish.
• Kurdish (Sorani dialect) remains significantly underexplored in ABSA research.

Purpose of the Study:

• To introduce the first publicly available ABSA dataset for the Sorani dialect of Kurdish.
• To address the critical gap in NLP resources for low-resource languages.
• To facilitate research and development in ABSA for Kurdish.

Main Methods:

• Creation of a dataset with over 4000 quadruplet ABSA entries from restaurant reviews in Kurdish (Sorani dialect).
• Utilized a prompt-based few-shot learning model for automated annotation of aspect-opinion-category-sentiment quadruples.
• Dataset annotation was guided by a manually verified support set, confirmed by native Kurdish speakers.

Main Results:

• Development of the first comprehensive ABSA dataset for Kurdish (Sorani dialect).
• The dataset contains >4000 quadruplet ABSA entries in the restaurant review domain.
• The dataset is annotated using the Perso-Arabic script.

Conclusions:

• This resource significantly advances NLP research for low-resource languages, specifically Kurdish.
• The dataset is suitable for training, fine-tuning, and benchmarking machine learning and deep learning models.
• It supports cross-lingual model adaptation and future ABSA research in Kurdish.