Evaluation of an Automatic Sign Language Production System for Breaking News
Abstract
Sign language production (SLP) systems using avatars have been proposed to quickly provide accessible information in sign language (SL) to Deaf and hard-of-hearing people during emergencies. However, generating accurate SL for breaking news remains difficult. This is due to two key challenges: the difficulty of translating text with a writing style that differs from general news, and the challenge of motion synthesis, which requires accurately representing important linguistic features such as spatial location, mouthing, and facial expressions. In this study, we constructed an SLP system by integrating a domain-specific translation model and motion control methods for signing location, mouthing, and facial expressions. We evaluated the system in a user study with 21 Deaf participants. The results showed that the integrated approach improved content comprehension and received higher legibility and naturalness ratings than the baseline conditions.