Analyzing the expression of hate speech in YouTube comments

Researchers at the Interactive Technologies Institute have conducted a study on hate speech targeting marginalized communities in YouTube comments. Cláudia Silva joined a research team led by Paula Carvalho (INESC-ID/IST), that aimed to analyze the prevalence and linguistic strategies underlying online comprehensively hate speech against Afro-descendant, Roma, and LGBTQ+ communities in Portugal. The research was developed within the scope of the project HATE COVID-19.PT.

The study, published in the Journal of Language Aggression and Conflict, was prompted by the increasing presence of Online Hate Speech (OHS) across social media platforms and its negative impact on individuals and society. However, the lack of a universal definition of hate speech and the absence of resources that can be easily operationalized and compared across different linguistic and pragmatic contexts posed significant challenges.

The researchers created a finely-grained annotated corpus called CO-HATE (Counter, Offensive and Hate speech) to address these limitations. This corpus comprises 20,590 comments posted by 8,485 unique online users on 39 YouTube videos targeting the aforementioned communities. The comments were meticulously analyzed using corpus linguistics techniques, allowing the team to identify overt and covert hate speech, counter-speech, and offensive speech patterns.

The study results provided valuable insights into the specificities of hate speech targeting Afro-descendant, Roma, and LGBTQ+ communities in Portugal. By combining quantitative and qualitative research methods, the researchers were able to shed light on the linguistic and rhetorical strategies employed in expressing hatred towards these marginalized groups. Furthermore, this research has led to the creation of the first finely-grained annotated corpus for European Portuguese, which will serve as a crucial resource for studying and detecting online hate speech, especially concerning the targeted communities, on social media platforms.

This study’s findings are expected to impact society in the medium and long term significantly. The insights gained from analyzing online hate speech will inform the development of more efficient automated detection models tailored to Portuguese. This will contribute to a deeper understanding of the dynamics of online hate speech in Portugal and aid in the fight against discrimination and intolerance. Additionally, the CO-HATE corpus will serve as an important tool for researchers and developers interested in detecting overt and covert hate speech and related phenomena, such as counter-speech and offensive speech, in Portuguese.

As for the future steps, the researchers are building upon the knowledge gained from this study to further their research in the context of the KNOwhate project. in which Cláudia Silva is leading a number of research activities related to countering hate speech in Portugal through media production (podcasts and an online campaign).

Several authors of this study also explore the relationship between hate speech, conspiracy theories, and social media in another project called MAICT. Overall, this research marks a significant step towards understanding and combating online hate speech. The study’s findings and resources will contribute to developing effective strategies to address hate speech, protect marginalized communities, and foster a more inclusive and tolerant digital environment.