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

Group imbalance, resulting from inadequate or unrepresentative datacollection methods, is a primary cause of representation bias in datasets.Representation bias can exist with respect to different groups of one or moreprotected attributes and might lead to prejudicial and discriminatory outcomestoward certain groups of individuals; in cases where a learning model istrained on such biased data. This paper presents MASC, a data augmentationapproach that leverages affinity clustering to balance the representation ofnon-protected and protected groups of a target dataset by utilizing instancesof the same protected attributes from similar datasets that are categorized inthe same cluster as the target dataset by sharing instances of the protectedattribute. The proposed method involves constructing an affinity matrix byquantifying distribution discrepancies between dataset pairs and transformingthem into a symmetric pairwise similarity matrix. A non-parametric spectralclustering is then applied to this affinity matrix, automatically categorizingthe datasets into an optimal number of clusters. We perform a step-by-stepexperiment as a demo of our method to show the procedure of the proposed dataaugmentation method and evaluate and discuss its performance. A comparison withother data augmentation methods, both pre- and post-augmentation, is conducted,along with a model evaluation analysis of each method. Our method can handlenon-binary protected attributes so, in our experiments, bias is measured in anon-binary protected attribute setup w.r.t. racial groups distribution for twoseparate minority groups in comparison with the majority group before and afterdebiasing. Empirical results imply that our method of augmenting dataset biasesusing real (genuine) data from similar contexts can effectively debias thetarget datasets comparably to existing data augmentation strategies.

Affinity Clustering Framework for Data Debiasing Using Pairwise Distribution Discrepancy