Hate Speech on Social Media






On demand

Credits: @Sunfwerr - https://twitter.com/Sunfwerr/status/1369291831549321217


The generation and spread of hate speech are part of a generalized phenomenon that affects the sharing and dissemination of content on social networks. The problems generated by this phenomenon are well known from the amplification of the general lack of trust on the part of society to the exacerbation of the mechanisms of social polarization or stigmatization of vulnerable groups. Reducing and understanding the generation and spread of hate speech is essential to develop well-founded policies and actions towards its detection and reduction of spread.

The technologies developed during this project aimed to understand and typify the end-to-end hate speech sparse process by applying state-of-the-art methodologies that allowed to characterize, in social networks, the massive generators of hate speech and the effects that these produced in the receptors. Towards this end, the project proposed to develop a new methodology that allows granting a trust and credibility score to both social media accounts and messages that may contain hate speech. To this end, it proposed the combination of computer vision and machine learning several techniques that allowed the extraction of a representation that combined text and image information, and the embedding of the message in a joint multimodal space.

This project was possible thanks to the Spanish Ministry of Economy and Competitiveness (MINECO) and the European Regional Development Fund (ERDF) under research project TIN2015-65464-R.