Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Yuejie, Yang, Ke, Hua, Yueying, Chen, Berlin, Nie, Jianhao, He, Yueping, Kang, Caixin
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2602.12783
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910215157317632
author Li, Yuejie
Yang, Ke
Hua, Yueying
Chen, Berlin
Nie, Jianhao
He, Yueping
Kang, Caixin
author_facet Li, Yuejie
Yang, Ke
Hua, Yueying
Chen, Berlin
Nie, Jianhao
He, Yueping
Kang, Caixin
contents Spoken query retrieval is an important interaction mode in modern information retrieval. However, existing evaluation datasets are often limited to simple queries under constrained noise conditions, making them inadequate for assessing the robustness of spoken query retrieval systems under complex acoustic perturbations. To address this limitation, we present SQuTR, a robustness benchmark for spoken query retrieval that includes a large-scale dataset and a unified evaluation protocol. SQuTR aggregates 37,317 unique queries from six commonly used English and Chinese text retrieval datasets, spanning multiple domains and diverse query types. We synthesize speech using voice profiles from 200 real speakers and mix 17 categories of real-world environmental noise under controlled SNR levels, enabling reproducible robustness evaluation from quiet to highly noisy conditions. Under the unified protocol, we conduct large-scale evaluations on representative cascaded and end-to-end retrieval systems. Experimental results show that retrieval performance decreases as noise increases, with substantially different drops across systems. Even large-scale retrieval models struggle under extreme noise, indicating that robustness remains a critical bottleneck. Overall, SQuTR provides a reproducible testbed for benchmarking and diagnostic analysis, and facilitates future research on robustness in spoken query to text retrieval.
format Preprint
id arxiv_https___arxiv_org_abs_2602_12783
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SQuTR: A Robustness Benchmark for Spoken Query to Text Retrieval under Acoustic Noise
Li, Yuejie
Yang, Ke
Hua, Yueying
Chen, Berlin
Nie, Jianhao
He, Yueping
Kang, Caixin
Information Retrieval
Artificial Intelligence
Spoken query retrieval is an important interaction mode in modern information retrieval. However, existing evaluation datasets are often limited to simple queries under constrained noise conditions, making them inadequate for assessing the robustness of spoken query retrieval systems under complex acoustic perturbations. To address this limitation, we present SQuTR, a robustness benchmark for spoken query retrieval that includes a large-scale dataset and a unified evaluation protocol. SQuTR aggregates 37,317 unique queries from six commonly used English and Chinese text retrieval datasets, spanning multiple domains and diverse query types. We synthesize speech using voice profiles from 200 real speakers and mix 17 categories of real-world environmental noise under controlled SNR levels, enabling reproducible robustness evaluation from quiet to highly noisy conditions. Under the unified protocol, we conduct large-scale evaluations on representative cascaded and end-to-end retrieval systems. Experimental results show that retrieval performance decreases as noise increases, with substantially different drops across systems. Even large-scale retrieval models struggle under extreme noise, indicating that robustness remains a critical bottleneck. Overall, SQuTR provides a reproducible testbed for benchmarking and diagnostic analysis, and facilitates future research on robustness in spoken query to text retrieval.
title SQuTR: A Robustness Benchmark for Spoken Query to Text Retrieval under Acoustic Noise
topic Information Retrieval
Artificial Intelligence
url https://arxiv.org/abs/2602.12783