The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes
This addresses the important problem of hate speech detection for online safety, but it is incremental as it introduces a new dataset rather than a novel method.
The authors tackled the problem of detecting hate speech in multimodal memes by creating a new challenge dataset with 'benign confounders' that make unimodal models struggle, and found that state-of-the-art methods achieve only 64.73% accuracy compared to humans at 84.7%.
This work proposes a new challenge set for multimodal classification, focusing on detecting hate speech in multimodal memes. It is constructed such that unimodal models struggle and only multimodal models can succeed: difficult examples ("benign confounders") are added to the dataset to make it hard to rely on unimodal signals. The task requires subtle reasoning, yet is straightforward to evaluate as a binary classification problem. We provide baseline performance numbers for unimodal models, as well as for multimodal models with various degrees of sophistication. We find that state-of-the-art methods perform poorly compared to humans (64.73% vs. 84.7% accuracy), illustrating the difficulty of the task and highlighting the challenge that this important problem poses to the community.