Sexism detection with semantic graphs

March 10th, 2022. 14:00 - 15:30

Sexism detection is an important task for online content moderation. Our goal is to identify sexist tweets in the 2021 EXIST dataset as precisely as possible. For this purpose we present a human-in-the-loop based rule learning method as well as a syntax and knowledge based graph representation. We utilize graph matching with semantic node similarity to match our rules to the graphs. We still have some work to do, but our results seem promising. We hope to have a discussion about our future plans and ideas.

Location: Zoom