SQuTR: A Robustness Benchmark for Spoken Query to Text Retrieval under Acoustic Noise

TL;DR

SQuTR is a comprehensive benchmark dataset and evaluation protocol designed to assess and improve the robustness of spoken query to text retrieval systems under various noisy acoustic conditions.

Contribution

It introduces a large-scale, multi-domain dataset with synthesized speech and environmental noise, enabling reproducible robustness evaluation for retrieval systems.

Findings

Performance drops significantly with increased noise levels.

Different retrieval systems exhibit varying robustness to noise.

Even large models struggle under extreme noise conditions.

Abstract

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.