Open Conference Systems, CLADAG2023

Font Size: 
Identification of misogynistic accounts on Twitter through Graph Convolutional Networks
Lara Fontanella, Emiliano del Gobbo

Last modified: 2023-07-16


Misogyny is characterized by feelings of hatred, dislike, and mistrust towards women simply because they are women, as well as ingrained prejudice against them. While a number of studies have concentrated on automatically detecting misogynistic content disseminated on various social media platforms, only a small number have considered the conduct and relationships of users who spread misogynistic comments. Our study centers on Twitter content due to the relatively straightforward availability of networked data. With a substantial collection of tweets at our disposal, we utilize both the text-based content and relational data depicted in the friend/follower network to jointly classify Twitter accounts based on a binary misogyny scheme. To achieve this, we employ Graph Convolutional Networks and evaluate the effectiveness of various embedding algorithms utilised to generate the feature matrix from the textual data.