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

Pretrained language models (PLMs), such as GPT2, have achieved remarkableempirical performance in text generation tasks. However, pretrained onlarge-scale natural language corpora, the generated text from PLMs may exhibitsocial bias against disadvantaged demographic groups. To improve the fairnessof PLMs in text generation, we propose to minimize the mutual informationbetween the semantics in the generated text sentences and their demographicpolarity, i.e., the demographic group to which the sentence is referring. Inthis way, the mentioning of a demographic group (e.g., male or female) isencouraged to be independent from how it is described in the generated text,thus effectively alleviating the social bias. Moreover, we propose toefficiently estimate the upper bound of the above mutual information viaimportance sampling, leveraging a natural language corpus. We also propose adistillation mechanism that preserves the language modeling ability of the PLMsafter debiasing. Empirical results on real-world benchmarks demonstrate thatthe proposed method yields superior performance in term of both fairness andlanguage modeling ability.

Toward Fairness in Text Generation via Mutual Information Minimization based on Importance Sampling