Category
A new tool designed to detect hidden biases in medical AI training data is particularly significant for African healthcare. With the continent's rapidly expanding medical AI…
Researchers have developed TTSYoruba, an advanced text-to-speech system specifically for the Yoruba language, a major language spoken across West Africa. This innovation addresses…
This study demonstrates a significant step in adapting advanced AI speech recognition technology for specific Kenyan languages, including Kikuyu, Dholuo, and Kalenjin. By…
The Durban University of Technology hosted the first African DataScientia International Symposium, launching a continental Living Lab focused on "Diversity-Aware Sovereign AI."…
This research significantly advances AI's capability to adapt multilingual models to low-resource languages by evaluating its novel DeltaMerge-LowRes method on four African…
Researchers have developed VEXMLM, an AI model specifically designed to improve natural language processing for Ge'ez-script languages like Amharic and Tigrinya, and 17 other…
AfriVox-v2 is a new benchmark specifically designed to evaluate speech recognition models under realistic and noisy African conditions. It focuses on African languages and diverse…
This research presents the first end-to-end Text-to-Speech (TTS) study for Efik, a tonal language spoken in Southeastern Nigeria, addressing its underrepresentation in digital…
A new study exploring Automatic Speech Recognition for low-resource African languages found that linguistic relatedness alone is not a reliable predictor for cross-lingual…
A new study reveals that low-resource African languages like Afrikaans, Kiswahili, isiXhosa, and isiZulu can be leveraged in multi-turn conversations to bypass safety features in…
The Nsanku benchmark specifically evaluates LLM translation capabilities for 43 Ghanaian languages, highlighting a critical gap in AI's performance for low-resource African…
A new AI technique, Crosslingual On-Policy Self-Distillation (COPSD), has been developed and proven to significantly enhance mathematical reasoning capabilities of large language…
This research develops a Setswana sentiment dataset and identifies temporal simultaneity as a critical factor for high-quality annotation. The findings offer valuable insights for…
A new jailbreak benchmark, TukaBench, has been developed for seven African languages to address the English-centric bias in Large Language Model safety evaluations. This research…
New research demonstrates that specialized AI models, fine-tuned on the WAXAL corpus of 19 African languages, significantly outperform larger general-purpose models in automatic…
This study reveals that simply increasing data volume does not guarantee improved AI performance for African languages, highlighting the need for language-sensitive dataset…
A new study investigates prompting strategies for Natural Language Inference (NLI) in low-resource African languages like Swahili, Yoruba, and Hausa. The research highlights that…
A new AI text-to-speech benchmark, OpenBibleTTS, includes 37 underrepresented languages, with specific models showing strong intelligibility and user preference in several African…
Dr. Shamsuddeen will discuss how large language models (LLMs) misrepresent African languages and cultural contexts, despite two decades of progress in AfricaNLP. He will highlight…
This research evaluates leading large language models for machine translation between English and two West African languages, Hausa and Fongbe. It highlights significant…
This research addresses the critical scarcity of digital text data for African languages by developing and evaluating ASR pipelines for Fongbe and Hausa. The project significantly…
A new study reveals a substantial "African Language Tax" in large language models, where African languages incur significantly higher costs and latency due to inefficient…
This research paper critically examines the licensing landscape of Natural Language Processing (NLP) corpora specifically for low-resource African languages. It reveals widespread…
Researchers have developed an open-source automatic speech recognition (ASR) system specifically for assessing children's reading in Bambara, a prominent African language. This…
