Speech datasets for Nigerian languages with almost no transcribed audio
Speech models fail on languages where very little transcribed audio exists, which covers most of what is actually spoken here.
Post an idea, however early. Other builders add evidence, challenge the thinking or take it a step further, so a rough thought grows into research anyone can pick up. Every idea has its own page you can share anywhere.
Speech models fail on languages where very little transcribed audio exists, which covers most of what is actually spoken here.
A paragraph is enough. Say what you are curious about and what you think might be true. Half formed is fine, that is the point.
People add what they have tried, link prior work, or challenge the assumption. Each thread becomes a small piece of open research.
Every idea has its own link for WhatsApp, Facebook and X, so the people who can answer it actually see it.