Voices of Fashion: 150-Hour Yoruba Speech Corpus for Egba Adire and Color Advisory Systems

Morayo Lawal Adunni, Olanrewaju Israel Lawal

Keywords: Automatic speech recognition (ASR), Multilingual NLP, Dataset curation & annotation, Yoruba speech dataset, Cultural heritage AI

Abstract:

The intersection of language, culture, and fashion offers a unique opportunity to design inclusive artificial intelligence (AI) systems that address underrepresented communities. This work focuses on the curation of a 150-hour Yoruba fashion speech dataset for the Egba people of Southwestern Nigeria, a subgroup of the Yoruba estimated at nearly 2 million speakers in Ogun State. The Egba are historically renowned for their mastery of Adire (tie-and-dye) textile craft, with Abeokuta recognized as the capital of Nigeria’s Adire industry and home to Itoku Market, the largest tie-and-dye marketplace in West Africa. Despite this cultural and economic significance, Egba communities remain largely excluded from AI-driven tools that could enhance creativity, advisory, and digital commerce in their local language.

We propose the collection of 150 hours of speech recordings involving 200+ Egba speakers across artisans, market women, designers, and young customers. The dataset will capture fashion-related expressions, fabric pattern terminologies, dyeing techniques, color combinations, and advisory dialogues in Yoruba with Egba dialectal variations. All recordings will be transcribed, annotated, and validated by Yoruba language experts to ensure both linguistic richness and cultural accuracy.

The resulting dataset will enable the development of localized AI applications, such as a Yoruba-speaking fashion advisory chatbot, capable of recommending designs, color blends, and tie-and-dye techniques in culturally grounded ways. Beyond advisory, the dataset supports automatic speech recognition (ASR), machine translation, and dialogue systems tailored to Yoruba-speaking populations. By situating this work at the intersection of low-resource language dataset curation, AI for cultural heritage, and digital empowerment, we aim to directly contribute to economic inclusion, cultural preservation, and sustainable innovation for one of West Africa’s most historically influential communities.

Submission Number: 57