Sign Recognition Technology: The Sem-Lex Benchmark
0
8
Sign language technology is crucial for creating more inclusive digital tools. Most modern technology—from voice assistants to text-based search engines—is designed around spoken or written languages, excluding sign language users. We want to change that.
In our study, "The Sem-Lex Benchmark: Modeling ASL Signs and Their Phonemes", by Lee Kezar, Elana Pontecorvo, Adele Daniels, Connor Baer, Ruth Ferster, Lauren Berger, Jesse Thomason, Zed Sevcikova Sehyr, and Naomi Caselli, we created a large dataset to help computers recognize ASL. This is a key step toward making technology more accessible to deaf communities.
Key Findings
We compiled a dataset of over 84,000 videos featuring 41 deaf signers. By training a computer to recognize signs from these videos, we’re working toward building technology that can understand and respond to ASL. This dataset allows for more accurate sign language recognition, which could transform how deaf people interact with digital tools—from using smartphones to navigating websites.
This project paves the way for ethically sourced, efficient, and reproducible sign language research and more successful sign recognition technologies down the line.
What This Means for the Future of Sign Language Technology
Our aim is to contribute to technology that better serves people who use ASL.The Sem-Lex Benchmark is a meaningful step toward building tools that can recognize and respond to ASL, helping make AI-enabled technology that acknowledges the diversity of human communication.
Read the full paper.
Related Posts
