Skip to page content

AEGLIX MIND

Connected intelligence.Security research, connected

AEGLIX news · Africa

Ubini Builds Edo Language Corpus Through Elder Recordings and XML Documentation

Based on reporting from Techcabal

Published by the source: Tue, 17 Nov 2026 13:44:59 +0000

AEGLIX News

In 2024, Stephanie Ohumu established Ubini following her grandmother's passing, which underscored how much unwritten Edo cultural knowledge disappears with elder speakers. The project has since conducted interviews with roughly 20 individuals and transcribed over 20 hours of recordings. Its focus remains on people aged 70 and above who use Edo as their primary language, aiming to preserve everyday dialogue, ceremonial speech, and specialized knowledge that is no longer commonly passed down.

Research published in 2026 by Kedir Hussen and six colleagues revealed that over 98% of African languages lack support in language models, highlighting the gap Ubini is working to close for Edo. Ohumu estimates the initiative needs a minimum of 250 participants, each contributing about two hours of recordings, to document a wider spectrum of speakers and knowledge. Every initial transcript is then subjected to quality assurance by two University of Benin professors who review and correct the text.

Ubini has created Mia Mia, a 27-minute documentary centered on Maria Evbagharu Osazee, an 85-year-old woman reflecting on her life and the erosion of cultural heritage. The film debuted at the Uncolonial Film Festival in Germany in August 2026 and was later chosen for the Human Screen Festival in Tunisia in November, with a Nigerian premiere planned for the Kano Indigenous Languages of Africa Film Market and Festival later that month. The completed recordings are structured into XML files that hold speaker turns, Edo transcriptions, English translations, and metadata about contributors and recording contexts.

Ubini's roadmap targets 50 hours of collected audio by December 2026, followed by 150 hours transcribed and translated by the close of 2027 with the help of 30 transcribers, building toward a 250-hour corpus by 2029 to form a structured dataset for language preservation and possible computational applications.

Report prepared 2026-11-17 · Based on reporting from Techcabal. Source publication date above; subsequent developments may not be reflected.

Sources

Read original source · Techcabal (opens in a new tab)

Privacy policy

Explore the story

Consult the cited publisher for the original reporting, updates and supporting evidence. Check the publication date and source before acting on a headline.

These are research tools; no CVE, indicator, or campaign relationship is inferred from this article.

More from Techcabal

← Back to security news