Audio moment retrieval improved by better feature extraction and detection
Overview and Meta-Analysis of DCASE 2026 Challenge Task 6: Audio Moment Retrieval from Long Audio
Sound
Summary
Audio moment retrieval means finding specific moments in a long audio recording based on a text query. The authors describe a challenge where teams tried to locate such moments using advanced computer methods. The results show that improving the way audio and text features are matched and detecting the right moment boundaries helped a lot. However, this problem is still difficult, and even the best systems only found about half of the correct moments.
What this means in practice
- •For audio content platforms: Improve search within long audio recordings by automatically finding relevant moments spoken about in user queries.$Commercial implications: Enables selling advanced audio search features to podcast and media streaming services to enhance user engagement.
- •For media monitoring teams: Locate precise time intervals in broadcast audio where specific topics or sounds occur to assist in content analysis.
Authors
Hokuto Munakata, Tatsuya Komatsu, Keisuke Imoto, Taichi Nishimura, Huang Xie, Tuomas Virtanen
Abstract
This paper presents an overview of the Detection and Classification of Acoustic Scenes and Events (DCASE) 2026 Challenge Task 6, Audio Moment Retrieval (AMR) from Long Audio. Given a several-minute-long audio recording and a free-form text query, AMR aims to retrieve temporal moments in the recording that match the query, where each moment is represented by a pair of start and end timestamps. This task requires effective cross-modal alignment and long-range temporal modeling. We describe the task definition, the evaluation metrics, the development and evaluation datasets, and a baseline system that combines a pre-trained MS-CLAP feature extractor with a Detection Transformer (DETR)-based moment-detection network. On the development data, the baseline trained on a manually annotated dataset and a synthetic dataset achieved Recall1@0.7 of 13.56%, indicating that AMR in long audio remains a challenging problem. The challenge attracted 21 teams, which submitted 59 systems in total. The three best systems achieved Recall1@0.7 of 48.59%, roughly 3.5 times the baseline score. The results show that strengthening the audio-text feature extractor and the moment-detection network led to substantial performance improvements. Furthermore, the top three teams boosted performance by applying confidence score calibration or ensembling across different temporal resolutions of features.