Nigerian-UK Startup Talksign Launches AI Data Marketplace to Bridge Sign Language Gap

Talksign, an AI startup with roots in Nigeria and the UK, has introduced a new marketplace designed to address the critical shortage of sign language data for artificial intelligence development. This platform allows Deaf communities, researchers, and companies to contribute, annotate, license, and access diverse sign language datasets, aiming to overcome a significant bottleneck that has hindered progress in sign language AI.

The core issue stems from the vast data asymmetry between spoken and signed languages. While speech recognition models benefit from decades of extensive web audio data, sign language corpora are typically small, limited in scope, and often not readily licensable. This lack of diverse and accessible data means that existing AI models struggle with regional variations, natural signing rhythms, and the nuanced facial grammar essential to sign languages.

The marketplace features four key functions: contribution, where signers are compensated for submitting data; annotation, where fluent signers are paid to label video data; licensing, providing clear terms and provenance; and access, offering a centralized hub for datasets across various languages and dialects. This initiative directly supports the development of more accurate and inclusive AI tools for the Deaf community.

Talksign itself has developed its own AI models, Palm 1.0 (speech-to-sign) and Echo 1.0 (sign-to-speech), boasting impressive semantic accuracy and structural similarity scores. The company's prior work includes Talksign-1, a bidirectional ASL translation tool. By launching this marketplace, Talksign is not only advancing its own product development but also fostering a collaborative ecosystem crucial for global sign language AI innovation, with a direct benefit to African Deaf communities.