A Contrastive Learning Approach to Mitigate Bias in Speech Models
This addresses fairness concerns in speech models for underperforming subgroups, though it is incremental as it builds on existing bias mitigation techniques.
The paper tackled performance imbalance across population subgroups in speech models by proposing a contrastive learning approach, which improved internal subgroup representations and reduced bias, as demonstrated on two spoken language understanding datasets and languages.
Speech models may be affected by performance imbalance in different population subgroups, raising concerns about fair treatment across these groups. Prior attempts to mitigate unfairness either focus on user-defined subgroups, potentially overlooking other affected subgroups, or do not explicitly improve the internal representation at the subgroup level. This paper proposes the first adoption of contrastive learning to mitigate speech model bias in underperforming subgroups. We employ a three-level learning technique that guides the model in focusing on different scopes for the contrastive loss, i.e., task, subgroup, and the errors within subgroups. The experiments on two spoken language understanding datasets and two languages demonstrate that our approach improves internal subgroup representations, thus reducing model bias and enhancing performance.