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

Aspect-Based Sentiment Analysis (ABSA) extends traditional sentiment analysis by not only identifying the overall sentiment of a text but also associating specific sentiments with deeper and granular insights. The main objective of ABSA is to accurately extract relevant aspects and determine the sentiment polarity associated with each. Although extensive research has been conducted on ABSA across various languages, low-resource languages such as Kurdish remain largely underexplored in this domain. To address this gap, the present study introduces the first publicly available aspect-based sentiment analysis dataset for the Sorani dialect of Kurdish, addressing a critical gap in natural language processing (NLP) research for low-resource languages. The dataset has >4000 quadruplet ABSA in the restaurant review domain, written in the Kurdish language (Sorani dialect) using the Perso-Arabic script. A prompt-based few-shot learning model was employed to automatically annotate the dataset with aspect-opinion-category-sentiment quadruples, guided by a manually annotated support set verified by native Kurdish-language experts. This resource is intended for use in machine learning, deep learning, and cross-lingual model adaptation, making it suitable for training, fine-tuning, and benchmarking.