Meaningful Pose-Based Sign Language Evaluation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Jiang, Zifan, Leong, Colin, Moryossef, Amit, Göhring, Anne, Rios, Annette, Cory, Oliver, Ivashechkin, Maksym, Tarigopula, Neha, Zhang, Biao, Sennrich, Rico, Ebling, Sarah
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914082256322560
author Jiang, Zifan
Leong, Colin
Moryossef, Amit
Göhring, Anne
Rios, Annette
Cory, Oliver
Ivashechkin, Maksym
Tarigopula, Neha
Zhang, Biao
Sennrich, Rico
Ebling, Sarah
author_facet Jiang, Zifan
Leong, Colin
Moryossef, Amit
Göhring, Anne
Rios, Annette
Cory, Oliver
Ivashechkin, Maksym
Tarigopula, Neha
Zhang, Biao
Sennrich, Rico
Ebling, Sarah
contents We present a comprehensive study on meaningfully evaluating sign language utterances in the form of human skeletal poses. The study covers keypoint distance-based, embedding-based, and back-translation-based metrics. We show tradeoffs between different metrics in different scenarios through automatic meta-evaluation of sign-level retrieval and a human correlation study of text-to-pose translation across different sign languages. Our findings and the open-source pose-evaluation toolkit provide a practical and reproducible way of developing and evaluating sign language translation or generation systems.
format Preprint
id arxiv_https___arxiv_org_abs_2510_07453
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Meaningful Pose-Based Sign Language Evaluation
Jiang, Zifan
Leong, Colin
Moryossef, Amit
Göhring, Anne
Rios, Annette
Cory, Oliver
Ivashechkin, Maksym
Tarigopula, Neha
Zhang, Biao
Sennrich, Rico
Ebling, Sarah
Computation and Language
We present a comprehensive study on meaningfully evaluating sign language utterances in the form of human skeletal poses. The study covers keypoint distance-based, embedding-based, and back-translation-based metrics. We show tradeoffs between different metrics in different scenarios through automatic meta-evaluation of sign-level retrieval and a human correlation study of text-to-pose translation across different sign languages. Our findings and the open-source pose-evaluation toolkit provide a practical and reproducible way of developing and evaluating sign language translation or generation systems.
title Meaningful Pose-Based Sign Language Evaluation
topic Computation and Language
url https://arxiv.org/abs/2510.07453