Sign Language AI

2 items

RESEARCHarXiv CS.CL·5/1/2026

Targeted Linguistic Analysis of Sign Language Models with Minimal Translation Pairs

The paper introduces ASL-MTP, a new benchmark dataset for analyzing how well sign language models capture linguistic phenomena and utilize multi-articulator cues. It uses this dataset to conduct a targeted linguistic analysis of a state-of-the-art ASL-to-English translation model.

machine learning Sign Language AI Benchmarking natural language processing

ARTICLEDEV.to AI·5/5/2026

We Built Sign Language AI for a Language With Almost No Dataset. Here's What That Actually Looks Like.

This article details the development of OmniSign, a real-time Lebanese Sign Language (LSL) translator, addressing the challenges of building AI for a language with an almost non-existent dataset. The author emphasizes that the hardest problems encountered were not technical but human. The inspiration came from witnessing communication difficulties between a deaf man and a barista in Beirut.

Dataset Low-Resource Language machine learning Sign Language AI