Every lab has to decide where to begin. We chose language and learning, because they sit underneath almost everything else we might want to build.
Language is infrastructure
A health assistant, a tool for a classroom, a service for a ministry: each one has to understand and speak the languages people actually use. For Somali and many other African languages, the speech data, text resources and evaluation sets that make this possible barely exist. Building them is slow, unglamorous work, and it is the work that unlocks the rest.
What we are studying
- Speech recognition and synthesis for Somali, including how dialect and everyday pronunciation affect accuracy.
- Conversational tutoring: how much feedback helps a learner, and when it gets in the way.
- Digital resources for underrepresented languages, and how to collect them responsibly.
- Evaluation with real learners, in real conditions, rather than only on benchmarks.
Understand the problem before choosing the technology.
What comes next
AfLearn is the first place this research reaches people. As the language work matures, it becomes the foundation for products in education, public services and the organisations we work with across East Africa.
