English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Speaker recognition is increasingly used in several everyday applicationsincluding smart speakers, customer care centers and other speech-drivenanalytics. It is crucial to accurately evaluate and mitigate biases present inmachine learning (ML) based speech technologies, such as speaker recognition,to ensure their inclusive adoption. ML fairness studies with respect to variousdemographic factors in modern speaker recognition systems are lagging comparedto other human-centered applications such as face recognition. Existing studieson fairness in speaker recognition systems are largely limited to evaluatingbiases at specific operating points of the systems, which can lead to falseexpectations of fairness. Moreover, there are only a handful of bias mitigationstrategies developed for speaker recognition systems. In this paper, wesystematically evaluate the biases present in speaker recognition systems withrespect to gender across a range of system operating points. We also proposeadversarial and multi-task learning techniques to improve the fairness of thesesystems. We show through quantitative and qualitative evaluations that theproposed methods improve the fairness of ASV systems over baseline methodstrained using data balancing techniques. We also present a fairness-utilitytrade-off analysis to jointly examine fairness and the overall systemperformance. We show that although systems trained using adversarial techniquesimprove fairness, they are prone to reduced utility. On the other hand,multi-task methods can improve the fairness while retaining the utility. Thesefindings can inform the choice of bias mitigation strategies in the field ofspeaker recognition.

To train or not to train adversarially: A study of bias mitigation strategies for speaker recognition