Ayoo: A Multilingual Multimodal Music Information Retrieval Dataset for Ghana Music Award Videos

Abstract Multimodal music information retrieval has gained significant momentum, producing a wide range of datasets for machine learning and deep learning tasks. However, these resources remain overwhelmingly skewed toward high-resource languages and Western musical traditions. To address this gap and support more culturally inclusive artificial intelligence, this article introduces \textit{Ayoo}, a multilingual and multimodal music information retrieval dataset derived from Ghana Music Awards videos. The dataset captures the linguistic and cultural diversity of Ghana across textual, visual, and audio modalities. Extracted from historical and contemporary award-winning performances spanning 1999 to 2023, Ayoo provides a structured foundation for downstream tasks such as genre classification, music sentiment analysis, emotion recognition, lyric--audio alignment, and culturally aware recommendation systems in low-resource multilingual settings. The article describes the dataset construction pipeline, annotation process, statistical properties, benchmarking applications, and principal limitations. Ayoo contributes a culturally grounded benchmark for music information retrieval and offers a reproducible framework for computational work on African music traditions. he source code and datasets have been made open access in a GitHub repository https://github.com/CAIRGH/Ayoo-Dataset