Over-Generation Cannot Be Rewarded: Length-Adaptive Average Lagging for Simultaneous Speech Translation

Sara Papi, Marco Gaido, Matteo Negri, Marco Turchi


Abstract
Simultaneous speech translation (SimulST) systems aim at generating their output with the lowest possible latency, which is normally computed in terms of Average Lagging (AL). In this paper we highlight that, despite its widespread adoption, AL provides underestimated scores for systems that generate longer predictions compared to the corresponding references. We also show that this problem has practical relevance, as recent SimulST systems have indeed a tendency to over-generate. As a solution, we propose LAAL (Length-Adaptive Average Lagging), a modified version of the metric that takes into account the over-generation phenomenon and allows for unbiased evaluation of both under-/over-generating systems.
Anthology ID:
2022.autosimtrans-1.2
Volume:
Proceedings of the Third Workshop on Automatic Simultaneous Translation
Month:
July
Year:
2022
Address:
Online
Venue:
AutoSimTrans
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
12–17
Language:
URL:
https://aclanthology.org/2022.autosimtrans-1.2
DOI:
10.18653/v1/2022.autosimtrans-1.2
Bibkey:
Cite (ACL):
Sara Papi, Marco Gaido, Matteo Negri, and Marco Turchi. 2022. Over-Generation Cannot Be Rewarded: Length-Adaptive Average Lagging for Simultaneous Speech Translation. In Proceedings of the Third Workshop on Automatic Simultaneous Translation, pages 12–17, Online. Association for Computational Linguistics.
Cite (Informal):
Over-Generation Cannot Be Rewarded: Length-Adaptive Average Lagging for Simultaneous Speech Translation (Papi et al., AutoSimTrans 2022)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-script-update/2022.autosimtrans-1.2.pdf
Code
 hlt-mt/fbk-fairseq