Entitled \u00abA systematic review of the literature on the assessment and mitigation of bias in automatic speech recognition models for low-resource African languages\u00bb, this study highlights a major challenge for inclusive artificial intelligence: biases related to gender, accent, dialect and linguistic under-representation, which significantly affect the performance of speech technologies in Africa.<\/p>\n\n\n\n
Conducted by Joyce Nakatumba-Nabende, Sulaiman Kagumire, Caroline Kantono and Peter Nabende (Makerere University, Uganda), this study presents a systematic review of the international scientific literature, following the PRISMA methodology, with the aim of identifying the main types of bias and the strategies employed to assess and reduce them.<\/p>\n\n\n\n
The authors show that, despite rapid advances in speech AI, African languages remain largely under-represented in research. Biases related to gender, accent and dialect are the most widely documented, whilst those related to age and ethnic origin remain almost entirely absent. The study also highlights that mitigation approaches remain predominantly data-centred (diversification, augmentation, transfer), with still limited use of advanced methods of responsible AI and fairness-aware modelling.<\/p>\n\n\n\n
This journal calls for greater engagement from the research community, policymakers and technology stakeholders to develop speech recognition systems capable of reflecting the linguistic and social diversity of the African continent, and to ensure that digital innovation benefits everyone.<\/p>\n\n\n\n
Authors: Joyce Nakatumba-Nabende, Sulaiman Kagumire, Caroline Kantono, Peter Nabende<\/strong><\/p>\n\n\n\n Read the full article below:<\/strong><\/p>\n\n\n\n