Artificial intelligence is advancing rapidly, but one area remains largely underserved: sign language.
While AI can now translate spoken languages, generate images and write code within seconds, sign languages have struggled to gain similar attention because of one major challenge — the lack of quality, accessible data.
African startup Talksign is working to change that.
Founded by Edidiong Ekong, Talksign has launched the Talksign Marketplace, a platform designed to help Deaf communities, researchers, institutions and technology companies contribute, annotate, license and access sign language datasets.
The goal is bigger than building another accessibility tool. Talksign wants to create the data infrastructure needed for AI systems that can properly understand and communicate through sign language.
Turning sign language data into an AI resource
Sign languages are full languages with their own grammar and structures. Unlike spoken language, communication happens through a combination of hand movements, facial expressions, body position and space.
This makes building AI models for sign language particularly challenging.
Talksign’s marketplace addresses this through four main areas: contribution, annotation, licensing and access.
Signers and Deaf organisations can contribute video data and receive compensation. The data can then be annotated and processed by fluent signers before being made available to researchers and AI companies through clear licensing terms.
This approach also creates a potential business model around data that has historically been difficult to access or commercialise.
Why Talksign matters to Africa’s AI ecosystem
The opportunity extends beyond sign language.
Many African languages, including local spoken and signed languages, remain poorly represented in AI because developers lack large, high-quality datasets.
Talksign’s model offers an interesting example of how African startups can turn this challenge into an opportunity by building data infrastructure around underserved communities while ensuring those communities benefit financially from the data they provide.
The company says its own AI models demonstrate what can become possible when better datasets are available.
Its Palm 1.0 speech-to-sign model reportedly achieves 84.2% semantic accuracy, while its Echo 1.0 sign-to-speech model reportedly operates at 30 frames per second.
Beyond accessibility, Talksign is showing that the next phase of Africa’s AI economy may not only be about building AI applications. It could also involve building the data, infrastructure and systems that allow those applications to work better.
For Business Verge, that’s the bigger story: Talksign is turning an overlooked problem into an AI infrastructure opportunity.
