Towards Machine Unlearning for Paralinguistic Speech Processing
Orchid Chetia, Girish*, Mohd Mujtaba, Shubham, Swarup Ranjan, Vandana, Muskaan, Arun Balaji, Rajesh · 2025
Read Towards Machine Unlearning for Paralinguistic Speech Processing by Orchid Chetia, Girish*, Mohd Mujtaba, Shubham, Swarup Ranjan, Vandana, Muskaan, Arun Balaji, Rajesh (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-02230.txt.
Description
Orchid Chetia Phukan and co-authors pioneer machine unlearning for speech paralinguistics in this 2025 paper. Unlearning makes a trained model forget chosen data, a need that grows as privacy rules tighten. The speech tasks here read feelings from voice: emotion recognition and depression detection.
Their method, SISA++, extends a leading approach that trains separate models on data shards and merges them by averaging weights. After unlearning on benchmark emotion and depression datasets, SISA++ kept more of its skill than the original method. The authors also give cookbook recipes, practical guidance on features and model shapes that soften the damage unlearning does, aiming to make the technique easier to adopt in real services.
Catalog record
| CREATOR | Orchid Chetia, Girish*, Mohd Mujtaba, Shubham, Swarup Ranjan, Vandana, Muskaan, Arun Balaji, Rajesh |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-02230