AMI 2018 Dataset

The AMI 2018 dataset collects 5,000 tweets written in Italian and 5,000 in English annotated for misogynistic content. Each instance is further annotated with information about the type of misogynistic behaviour (stereotype, objectification, dominance, derailing, sexual harassment, threats of violence, discredit) and target category (individual or generic target).<p> The dataset has been used in the AMI shared task (https://amievalita2018.wordpress.com) during the Evalita 2018 evaluation campaign (http://www.evalita.it/2018). The task challenged participants to build systems that automatically identify misogynous content in Twitter.<p>In order to comply with GDPR privacy rules and Twitter’s policies, the identifiers of tweets and users have been anonymized and replaced by unique identifiers. <p>