Text-to-Audio Grounding: Building Correspondence Between Captions and Sound Events

TL;DR

This paper introduces the Text-to-Audio Grounding task and dataset, aiming to link captions with specific sound events in audio clips, advancing cross-modal audio understanding.

Contribution

It presents a new dataset and task for grounding captions in audio, bridging audio processing and language understanding.

Findings

Developed the AudioGrounding dataset with sound event locations.

Proposed a baseline approach achieving 28.3% event-F1 score.

Achieved a 14.7% PSDS score on the task.

Abstract

Automated Audio Captioning is a cross-modal task, generating natural language descriptions to summarize the audio clips' sound events. However, grounding the actual sound events in the given audio based on its corresponding caption has not been investigated. This paper contributes an AudioGrounding dataset, which provides the correspondence between sound events and the captions provided in Audiocaps, along with the location (timestamps) of each present sound event. Based on such, we propose the text-to-audio grounding (TAG) task, which interactively considers the relationship between audio processing and language understanding. A baseline approach is provided, resulting in an event-F1 score of 28.3% and a Polyphonic Sound Detection Score (PSDS) score of 14.7%.