Towards Machine Unlearning for Paralinguistic Speech Processing

TL;DR

This paper introduces SISA++, an improved machine unlearning method for paralinguistic speech tasks like emotion recognition and depression detection, demonstrating better performance retention and providing practical guidelines for future research.

Contribution

We propose SISA++, an enhanced machine unlearning technique for PSP tasks, and offer actionable recommendations to facilitate adoption and reduce performance loss.

Findings

SISA++ outperforms SISA in preserving performance after unlearning.

SISA++ achieves better results on CREMA-D and E-DAIC datasets.

Guidelines are provided for selecting features and architectures to improve unlearning efficacy.

Abstract

In this work, we pioneer the study of Machine Unlearning (MU) for Paralinguistic Speech Processing (PSP). We focus on two key PSP tasks: Speech Emotion Recognition (SER) and Depression Detection (DD). To this end, we propose, SISA++, a novel extension to previous state-of-the-art (SOTA) MU method, SISA by merging models trained on different shards with weight-averaging. With such modifications, we show that SISA++ preserves performance more in comparison to SISA after unlearning in benchmark SER (CREMA-D) and DD (E-DAIC) datasets. Also, to guide future research for easier adoption of MU for PSP, we present ``cookbook recipes'' - actionable recommendations for selecting optimal feature representations and downstream architectures that can mitigate performance degradation after the unlearning process.