HATE COVID-19 – Detecting Overt and Covert Hate Speech in Social Media
This project seeks to contribute to the analysis and detection of online hate speech in Portuguese,
investigating the linguistic and rhetorical strategies underlying both overt and covert hatred content.
Specifically, the key objectives of this project are twofold. Firstly, it attempts to develop methods for
creating a large-scale Portuguese annotated corpus covering both overt and covert online hate speech,
before and during the Covid-19 pandemic. Secondly, it intends to develop a machine learning prototype
that demonstrates how the information in the annotated corpus can support hate speech detection,
considering attributes like hate speech target, and highlighting the linguistic and rhetorical clues
underlying hatred content.
Based on a qualitative approach, this article explores hate speech through the eyes of the most representative minorities in Portugal, namely the Afro-descendant, Roma and LGBTQ+ communities. To this end, we conducted three focus groups, one with each community (n=17), which allowed us to investigate how the target groups perceive and experience hate speech in the Portuguese social and geopolitical context and to identify the most harmful forms of speech from their own perspective. Our results show that covert hate speech is more harmful than overt hate speech. More insightful, we found that those covert forms of hate speech often manifest as compliments and humour. These communicative strategies, anchored in positive or negative stereotypes, have motivated the normalization of hate speech, which justifies the research of strategies that allow their identification, namely in the context of automatic detection of online hate speech.
01/05/2021 - 30/07/2022
Fundação para a Ciência e Tecnologia
LUSA, Centro Nacional de Cibersegurança
Blending Interaction: Engineering Interactive Systems & Tools; Critical Computing; Learning, Education, and Families; Specific Applications Areas; Sustainability, and Social Justice; Understanding People