ArSL Friday Sermon Dataset

A dataset containing labeled Arabic Sign Language (ArSL) videos has been created to study the translation of speech and text into sign language, focusing on enhancing deaf individuals' comprehension of formal speech related to religious topics such as Friday sermons. The dataset comprises 262 videos featuring 131 phrases or words performed by two expert signers and annotated using the ELAN program with glossary words. Each video pair for every phrase or word was assessed using two feature extraction methods (MobileNetV2 and MediaPipe frameworks), followed by cosine similarity calculations and Euclidean distance. Additionally, Jaccard similarities were employed to analyze the overall gloss discrepancies between the signers' videos. These resources serve as valuable assets for future research in ArSL translation and technological advancements.