Whispering Across the Continent: Collecting and Analyzing African Culture Using Community Radios

joseph Domguia, Guy Adingono Nkama

Keywords: NLP, Text to Speech, Culture, African culture, Community radio, Oral tradition, Cultural preservation, Data collection, Whisper model, Multilingual, Language model, African language, Heritage, Audio data, Transcription, Translation, Cameroon, Social impact

TL;DR: Recording and Analyzing African Culture Using Community Radios

Abstract: African culture is rich and diverse, but much of its knowledge is held by the elders of the community and passed down through oral traditions. With globalization, young Africans are becoming increasingly disconnected from their roots, making it essential to collect and preserve this knowledge. However, the lack of accessible data on African culture presents a significant challenge. This research aims to address this problem by exploring new ways to collect and preserve African cultural data. Specifically, we have developed a device to perform continuous recording of cultural radio programs in local languages, which has enabled us to collect over 1500 hours of audio data. We are also exploring the use of a whisper model, which has proven to be effective in outperforming human transcription and being multilingual. This research project's final goal is to build a language model that understands African culture, providing an effective approach to store this knowledge for future generations to learn about their culture.