New Thiomi Dataset provides over 601,000 text annotations and 385,000 audio recordings across ten African languages to advance low-resource language AI models.

Dataset covers ten African languages including Swahili, Kikuyu, Wolof, Somali, and Fulani across four language families, collected through a community platform with over 100 contributors

Achieves 86-100% text approval rates through multi-tier quality assurance pipeline for six primary languages

Establishes baselines for ASR, machine translation, and text-to-speech models, with best ASR system reaching 3.24% word error rate on Swahili—significantly improving prior academic performance from 8% WER