Benchmarking Time-localized Explanations for Audio Classification Models

TL;DR

This paper introduces a benchmark for evaluating time-localized explanations in audio classification models using event annotations as ground truth, enabling systematic comparison and optimization of explanation methods.

Contribution

It proposes a novel benchmark for time-localized explanations in audio models based on event annotations, facilitating evaluation and improvement of explanation techniques.

Findings

Some explanation methods achieve near-perfect accuracy.

The benchmark helps identify spurious correlations in models.

Systematic optimization improves explanation quality.

Abstract

Most modern approaches for audio processing are opaque, in the sense that they do not provide an explanation for their decisions. For this reason, various methods have been proposed to explain the outputs generated by these models. Good explanations can result in interesting insights about the data or the model, as well as increase trust in the system. Unfortunately, evaluating the quality of explanations is far from trivial since, for most tasks, there is no clear ground truth explanation to use as reference. In this work, we propose a benchmark for time-localized explanations for audio classification models that uses time annotations of target events as a proxy for ground truth explanations. We use this benchmark to systematically optimize and compare various approaches for model-agnostic post-hoc explanation, obtaining, in some cases, close to perfect explanations. Finally, we illustrate the utility of the explanations for uncovering spurious correlations.