UNICEF reports that while the use of the internet is beneficial in many ways, it also creates the possibility of serious risks such as child-targeted personal data collection, cyberbullying and other forms of peer-to-peer violence, as well as exposure of children to hate speech and violent content.

NEIKEA aims to redefine what hate speech means for children, and how it is experienced from their perspective. Peer-to-peer teasing, offensive language, or exclusion can have a profound impact on young people, and understanding this is essential for effective prevention.

The project will address both hate speech directed at children and peer-to-peer online harm. NEIKEA collects and analyses linguistic data to capture how children offend and are offended, enriching resources with child-centred vocabulary. This perspective-based approach will support the prevention of cyberbullying and other forms of online harm.

To protect children from exposure to online hate, the project aims to develop an innovative hate speech radar. NEIKEA also emphasizes education and awareness. Through campaigns, printed dictionaries, and performative activities, children will be encouraged to recognise what is offensive and how to engage in respectful dialogue. By combining technology, research, and awareness-raising, NEIKEA intends to foster safer, more inclusive digital spaces for the next generation.

References

Poletto et al. (2021) – Resources and benchmark corpora for hate speech detection: a systematic review.
Systematic review of computational approaches to hate speech analysis. (Language Resources and Evaluation)

Caselli et al. (2020) – HateBERT: Retraining BERT for abusive language detection in English.
First language model specifically trained for hate speech detection. (Proceedings of the 5th Workshop on Online Abuse and Harms, WOAH 2021)

Basile et al. (2019) – SemEval-2019 Task 5: Multilingual detection of hate speech against immigrants and women in Twitter.
International campaign evaluating computational approaches to multilingual hate speech analysis. (Proceedings of the 13th International Workshop on Semantic Evaluation)

Bassignana et al. (2018) – HurtLex: A multilingual lexicon of words to hurt.
Multilingual computational lexicon of offensive and hateful expressions. (Proceedings of the Fifth Italian Conference on Computational Linguistics, CLiC-it 2018)

Pamungkas et al. (2021) – A joint learning approach with knowledge injection for zero-shot cross-lingual hate speech detection.
Innovative hybrid methodology combining lexical resources and neural language models for hate speech analysis. (Information Processing & Management)