Arabic Multimodal Dataset for Aspect-Based Sentiment Analysis
A new Arabic MABSA dataset is constructed and annotated, comprising 1721 Arabic hotel reviews associated with 3653 images collected from Booking.com, and a human annotation pipeline with well-defined annotation guidelines is developed, structured into three stages adopted from prior MABSA research.