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/

Zendy Plus

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

Zendy Tools

Zendy Open

Optimizing NLP models for fairness poses many challenges. Lack ofdifferentiable fairness measures prevents gradient-based loss training orrequires surrogate losses that diverge from the true metric of interest. Inaddition, competing objectives (e.g., accuracy vs. fairness) often requiremaking trade-offs based on stakeholder preferences, but stakeholders may notknow their preferences before seeing system performance under differenttrade-off settings. To address these challenges, we begin by formulating adifferentiable version of a popular fairness measure, Accuracy Parity, toprovide balanced accuracy across demographic groups. Next, we show howmodel-agnostic, HyperNetwork optimization can efficiently train arbitrary NLPmodel architectures to learn Pareto-optimal trade-offs between competingmetrics. Focusing on the task of toxic language detection, we show thegenerality and efficacy of our methods across two datasets, three neuralarchitectures, and three fairness losses.

Finding Pareto Trade-offs in Fair and Accurate Detection of Toxic Speech