WeSign — AI-Powered Real-Time Feedback for Vietnamese Sign Language Learning, Hanoi

Vietnam's Ministry of Education and Training has issued national standards for Vietnamese Sign Language (VSL) instruction, but teachers of deaf and hard-of-hearing pupils often lack the fluency to model precise fingerspelling and sign production, and self-directed practice at home is rare. Researchers from Hanoi University of Science and Technology, Hanoi National University of Education, and Hanoi Metropolitan University built WeSign to address this gap.

To give deaf and hard-of-hearing primary pupils real-time, AI-scored feedback on fingerspelling and sign formation via webcam, supporting VSL learning aligned with the Ministry's national standards.

First-grade lesson content was co-developed with teachers at a specialised school for deaf children in Hanoi, aligned to the Ministry's Circular on national VSL standards. The team ran a mixed-methods evaluation: a needs-assessment survey of 30 educators across three specialised institutions, followed by a pilot with 30 participants (teachers and pupils) at one specialised Hanoi school.

System Usability Scale results averaged 92.58 out of 100 ('best imaginable'), with teachers rating it 94.00 and students 92.30. A related paper by an overlapping research team benchmarked the underlying fingerspelling recognition model at over 91% accuracy on raw video and 95% on processed data across 23 VSL alphabet signs and 2 accent marks.

The evidence base is still early: the classroom pilot measured usability and acceptance rather than measured learning gains, involved a single school and 30 participants, and the published sources do not detail data-protection or consent safeguards for the video capture the recognition feature depends on.