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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Hate speech is an important problem in the management of user-generatedcontent. To remove offensive content or ban misbehaving users, contentmoderators need reliable hate speech detectors. Recently, deep neural networksbased on the transformer architecture, such as the (multilingual) BERT model,achieve superior performance in many natural language classification tasks,including hate speech detection. So far, these methods have not been able toquantify their output in terms of reliability. We propose a Bayesian methodusing Monte Carlo dropout within the attention layers of the transformer modelsto provide well-calibrated reliability estimates. We evaluate and visualize theresults of the proposed approach on hate speech detection problems in severallanguages. Additionally, we test if affective dimensions can enhance theinformation extracted by the BERT model in hate speech classification. Ourexperiments show that Monte Carlo dropout provides a viable mechanism forreliability estimation in transformer networks. Used within the BERT model, itofers state-of-the-art classification performance and can detect less trustedpredictions. Also, it was observed that affective dimensions extracted usingsentic computing methods can provide insights toward interpretation of emotionsinvolved in hate speech. Our approach not only improves the classificationperformance of the state-of-the-art multilingual BERT model but the computedreliability scores also significantly reduce the workload in an inspection ofofending cases and reannotation campaigns. The provided visualization helps tounderstand the borderline outcomes.

To BAN or not to BAN: Bayesian Attention Networks for Reliable Hate Speech Detection