Better Late Than Never: Meta-Evaluation of Latency Metrics for Simultaneous Speech-to-Text Translation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Polák, Peter, Papi, Sara, Bentivogli, Luisa, Bojar, Ondřej
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915838872780800
author Polák, Peter
Papi, Sara
Bentivogli, Luisa
Bojar, Ondřej
author_facet Polák, Peter
Papi, Sara
Bentivogli, Luisa
Bojar, Ondřej
contents Simultaneous speech-to-text translation systems must balance translation quality with latency. Although quality evaluation is well established, latency measurement remains a challenge. Existing metrics produce inconsistent results, especially in short-form settings with artificial presegmentation. We present the first comprehensive meta-evaluation of latency metrics across language pairs and systems. We uncover a structural bias in current metrics related to segmentation. We introduce YAAL (Yet Another Average Lagging) for a more accurate short-form evaluation and LongYAAL for unsegmented audio. We propose SoftSegmenter, a resegmentation tool based on soft word-level alignment. We show that YAAL and LongYAAL, together with SoftSegmenter, outperform popular latency metrics, enabling more reliable assessments of short- and long-form simultaneous speech translation systems. We implement all artifacts within the OmniSTEval toolkit: https://github.com/pe-trik/OmniSTEval.
format Preprint
id arxiv_https___arxiv_org_abs_2509_17349
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Better Late Than Never: Meta-Evaluation of Latency Metrics for Simultaneous Speech-to-Text Translation
Polák, Peter
Papi, Sara
Bentivogli, Luisa
Bojar, Ondřej
Computation and Language
Artificial Intelligence
Simultaneous speech-to-text translation systems must balance translation quality with latency. Although quality evaluation is well established, latency measurement remains a challenge. Existing metrics produce inconsistent results, especially in short-form settings with artificial presegmentation. We present the first comprehensive meta-evaluation of latency metrics across language pairs and systems. We uncover a structural bias in current metrics related to segmentation. We introduce YAAL (Yet Another Average Lagging) for a more accurate short-form evaluation and LongYAAL for unsegmented audio. We propose SoftSegmenter, a resegmentation tool based on soft word-level alignment. We show that YAAL and LongYAAL, together with SoftSegmenter, outperform popular latency metrics, enabling more reliable assessments of short- and long-form simultaneous speech translation systems. We implement all artifacts within the OmniSTEval toolkit: https://github.com/pe-trik/OmniSTEval.
title Better Late Than Never: Meta-Evaluation of Latency Metrics for Simultaneous Speech-to-Text Translation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2509.17349