Partner · evaluate
Initiate a partnership.
BantuNomics Stories licenses page-length connected-speech Bantu data across three modes — Bantu only, Bantu & English code-switch, and Bantu-accented English — plus the accent ASR benchmark and the reproduce-the-WER eval set. Consented, de-identified, and growing. Evaluate it on your own models, then license the catalogue.
Tell us about your evaluation
Start the conversation.
What you'd be licensing
Start with the public proof.
The three modesWhy connected speech comes in three flavours — and how each is labelled.
The coverage matrixEvery language on the same frame — what's filled, what's coming.
Accent ASR benchmarkHow a frontier ASR handles Bantu-accented English, measured.
One recording, six datasetsWhy this data is infrastructure, not a dataset.