Exploring Pose-based Sign Language Translation: Ablation Studies and Attention Insights

Tomáš Železný, Jakub Straka, Václav Javorek, Ondřej Valach, Marek Hrúz, Ivan Gruber · 2025

Read Exploring Pose-based Sign Language Translation: Ablation Studies and Attention Insights by Tomáš Železný, Jakub Straka, Václav Javorek, Ondřej Valach, Marek Hrúz, Ivan Gruber (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-01532.txt.

Description

Tomas Zelezny and five co-authors studied how to turn sign language into written text, in a 2025 paper. Their system works from pose data, the tracked points of the body and hands, rather than from raw video. They used a changed form of a known text model to read those poses.

Working with two public video sets, they tested ways to clean up the pose data before use. Scaling the data to a common size, filling gaps, and adding varied copies all helped the system work better on new material. A close look at where the model pays attention also showed a gain from adding one extra marker token. Code and cleaned data were shared in public.

Catalog record

CREATORTomáš Železný, Jakub Straka, Václav Javorek, Ondřej Valach, Marek Hrúz, Ivan Gruber
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-01532

Browse the full library · About the Tower of Babel