Bypassing the nuances of Portuguese covert hate speech through contextual analysis

dc.contributor.authorRamos, G.
dc.contributor.authorBatista, F.
dc.contributor.authorRibeiro, R.
dc.contributor.authorFialho, P.
dc.contributor.authorMoro, S.
dc.contributor.authorFonseca, A.
dc.contributor.authorGuerra, R.
dc.contributor.authorCarvalho, P.
dc.contributor.authorMarques, C.
dc.contributor.authorSilva, C.
dc.contributor.editor Santos, Manuel Filipe
dc.contributor.editorMachado, José
dc.contributor.editorNovais, Paulo
dc.contributor.editorCortez, Paulo
dc.contributor.editorMoreira, Pedro Miguel
dc.date.accessioned2026-11-17T09:56:53Z
dc.date.embargo2025-11-16
dc.date.issued2025
dc.date.updated2026-11-17T09:54:00Z
dc.description.abstractDetecting and addressing covert hate speech poses significant challenges in online platforms where discriminatory messages are often disguised within seemingly innocuous content. In this study, we investigate the effectiveness of contextual analysis in bypassing the nuances of covert hate speech. Our research explores the impact of prompt engineering and context addition on the classification of overt and covert hate speech across diverse target groups, including Roma, migrants, LGBTQ+, and individuals of African descent. Through experimental trials using generative models like GPT-3.5 and GPT-4, our findings reveal that the addition of context, not only improves the overall performance of the models in general hate speech (from 75.0% to 79.6% F1 score, for the positive class), but also significantly improves the classification of covert hate speech, increasing True Positives by 21.64% (absolute) compared to the 6.5% in overt hate speech. Despite these improvements, the addition of context also increased the number of False Positives, indicating that a further refinement is needed for this contextual analysis. Moreover, target group analysis demonstrates a correlation between the prevalence of covert hate speech and model performance.eng
dc.event.date2024
dc.event.locationViana do Casteloeng
dc.event.title23rd EPIA Conference on Artificial Intelligence
dc.event.typeConferênciapt
dc.identifier.citationRamos, G., Batista, F., Ribeiro, R., Fialho, P., Moro, S., Fonseca, A., Guerra, R., Carvalho, P., Marques, C., & Silva, C. (2025). Bypassing the nuances of Portuguese covert hate speech through contextual analysis. In M. F. Santos, J. Machado, P. Novais, P. Cortez, & P. M. Moreira (Eds.), Progress in artificial intelligence: EPIA 2024 (pp. 241-253). Springer. https://doi.org/10.1007/978-3-031-73503-5_20
dc.identifier.doi10.1007/978-3-031-73503-5_20
dc.identifier.isbn978-3-031-73503-5
dc.identifier.issn0302-9743
dc.identifier.urihttps://hdl.handle.net/10071/38862
dc.language.isoeng
dc.pagination241 - 253
dc.peerreviewedyes
dc.publisherSpringer
dc.relationCERV-2021-EQUAL, 101049306
dc.relation.ispartofProgress in artificial intelligence: EPIA 2024
dc.rightsembargoedAccess
dc.subjectGPTeng
dc.subjectHate speecheng
dc.subjectNatural language processingeng
dc.subject.fosDomínio/Área Científica::Ciências Naturais::Ciências da Computação e da Informaçãopor
dc.subject.fosDomínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informáticapor
dc.subject.fosDomínio/Área Científica::Ciências Sociais::Ciências da Comunicaçãopor
dc.subject.fosDomínio/Área Científica::Ciências Sociais::Outras Ciências Sociaispor
dc.subject.fosDomínio/Área Científica::Humanidades::Línguas e Literaturaspor
dc.titleBypassing the nuances of Portuguese covert hate speech through contextual analysiseng
dc.typeconferenceObject
dc.volume14969 LNAI
iscte.alternateIdentifiers.scopus2-s2.0-85210438856
iscte.identifier.cienciahttps://ciencia.iscte-iul.pt/id/ci-pub-106476
iscte.subject.odsPaz, justiça e instituições eficazespor

Ficheiros

Pacote original

A mostrar 1 - 1 de 1
A carregar...
Nome:
conferenceObject_106476.pdf
Tamanho:
366.95 KB
Formato:
Adobe Portable Document Format
Descrição:
Versão Aceite