Tag
Tether's AI Research group has developed TranslatePsy-AfriSLM, a suite of open-source machine translation models covering 19 Sub-Saharan African languages. This initiative aims to…
This research directly addresses a critical methodological challenge in building AI for African languages, showing that current synthetic data quality metrics do not predict…
AfriSyCo, a new research study, examines how AI language models respond to factual content in African languages, revealing that assertive framing significantly increases the…
Tether AI Research has released open-source, offline translation models for 19 African languages, designed to run on basic smartphones without internet access. This addresses…
Lelapa AI, a South African company, is focused on building efficient AI for African languages, developing speech-to-text and text-to-speech solutions for enterprise use. Their…
South African AI company Lelapa AI is adding text-to-speech to its product line, aiming to enable devices to speak in vernacular languages across Africa. The firm is also looking…
AfriSwitch is a new 61.36-hour benchmark dataset of human-transcribed, real-world code-switched speech across 16 African languages. It reveals that current AI speech recognition…
Intron's Sahara v2.5 significantly advances African voice AI by enabling computers to understand and process conversations that mix multiple languages. This innovation is crucial…
Lagos-based Intron has unveiled Sahara v2.5, a voice AI system specifically designed to handle code-switching in African languages, a common linguistic pattern across the…
TranslatePsy-AfriSLM introduces open-source machine translation resources for 19 Sub-Saharan African languages, including curated and synthetic data, and fine-tuned SLMs. These…
A new open-source project, TranslatePsy-AfriSLM, is tackling the critical lack of high-quality machine translation for 19 Sub-Saharan African languages. This initiative provides…
This study critically examines adaptive inference for Natural Language Inference (NLI) in 15 African languages, highlighting issues with the AfriXNLI benchmark's data overlap and…
The LINGUA Africa initiative is funding 26 projects to develop AI resources and applications for over 50 African languages, dialects, and sign languages across 47 countries. These…
A new study demonstrates that AI safety measures, primarily built in English, are largely ineffective in African languages like Twi, Hausa, Amharic, and Swahili. This exposes a…
The ATLAS Umoja AI initiative has been launched by African governments and technology partners to accelerate AI development in African languages, aiming to embed local cultures…
DONDO is an open-source AI project providing automatic speech recognition models specifically for twenty-seven African language varieties across six countries. These models aim to…
This research directly addresses the challenges of developing automatic speech recognition for low-resource African languages. The study utilized two Africa-centric datasets to…
DONDO is an open-source initiative providing new AI-powered automatic speech recognition models specifically for twenty-seven African language varieties across Ghana, Sierra…
This study demonstrates a significant step in adapting advanced AI speech recognition technology for specific Kenyan languages, including Kikuyu, Dholuo, and Kalenjin. By…
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…
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…
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…
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…
