October 16, 2026 -- After months of back-and-forth between our team and our institutional IRB, we have finally been granted IRB approval to begin building our database of African voices. This is the first step in a long journey to respected and robust language technology for African languages. Beginning with investment and research into African tonal and stress-aligned languages, we hope to pioneer a new frontier in African linguistics and machine learning.
For much too long, has tone, a crucial facet of meaning in many African languages, been the victim of being quietly ignored. Automated speech recognition systems designed to be "multilingual" are missing this crucial element in their training datasets. How can a system designed to be language agnostic claim to be so and neglect the one element that defines meaning in two-fifths of the world's languages? With ALTER, we plan to research implementations of speech technology using painstakingly-annotated data, the first of its kind, to expand what is possible for African languages that was originally not possible in its past.
As we begin our protocol and begin working with participants, we turn to the community. We have no intention of going about this alone because community is what defines language work and it is the only way to ensure that every voice is heard. We are looking for representatives or community members who would be interested in collaborating with us to contribute existing work to this database and establish institutional relationships and make it a testament to masakhane (we build together) and what it means. We look forward to learning more about what you have and what we can do together.
This message is not a formal recruitment message for participants in our IRB-approved study or for new data collection, we seek only professional or institutional relationships for futher collaborations.