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

With the vast development and employment of artificial intelligenceapplications, research into the fairness of these algorithms has beenincreased. Specifically, in the natural language processing domain, it has beenshown that social biases persist in word embeddings and are thus in danger ofamplifying these biases when used. As an example of social bias, religiousbiases are shown to persist in word embeddings and the need for its removal ishighlighted. This paper investigates the state-of-the-art multiclass debiasingtechniques: Hard debiasing, SoftWEAT debiasing and Conceptor debiasing. Itevaluates their performance when removing religious bias on a common basis byquantifying bias removal via the Word Embedding Association Test (WEAT), MeanAverage Cosine Similarity (MAC) and the Relative Negative Sentiment Bias(RNSB). By investigating the religious bias removal on three widely used wordembeddings, namely: Word2Vec, GloVe, and ConceptNet, it is shown that thepreferred method is ConceptorDebiasing. Specifically, this technique manages todecrease the measured religious bias on average by 82,42%, 96,78% and 54,76%for the three word embedding sets respectively.

"Thy algorithm shalt not bear false witness": An Evaluation of Multiclass Debiasing Methods on Word Embeddings