Not all comments are equal: Insights into comment moderation from a topic-aware model

Moderation of reader comments is a signifi- cant problem for online news platforms. Here, we experiment with models for automatic mod- eration, using a dataset of comments from a popular Croatian newspaper. Our analy- sis shows that while comments that violate the moderation rules mostly share common linguistic and thematic features, their con- tent varies across the different sections of the newspaper. We therefore make our models topic-aware, incorporating semantic features from a topic model into the classification de- cision. Our results show that topic informa- tion improves the performance of the model, increases its confidence in correct outputs, and helps us understand the model’s outputs.