Detecting Misogynia in Italian Tweets: A Linguistic and Contextual based Method

Social media provides a platform for people to freely share their lives and feelings. However, with the increase of people's use of social media, the dark side of social media has gradually attracted people's attention. The anonymity of social media has led many people to use it to harass, abuse, and attack others, with women being a common target group. Twitter, one of the world's most well-known social media platforms, is used by hundreds of millions of people worldwide to communicate and share information every day. Although the artificial intelligence system for detecting hate comments has been applied to the maintenance of social media, it is still difficult to distinguish hate speech against women. That is because misogynistic comments are often subtle, and the currently commonly applied detection methods do not sufficiently consider the contextual information of the text. Therefore, we propose a BERT-based misogyny detection method that can fully integrate contextual information and context to classify texts. The experimental results show that our method has significant improvement compared with other methods.