DyCR-Net: A dynamic context-aware routing network for multi-modal sarcasm detection in conversation

Xingjie Zhuang, Zhixin Li, Fengling Zhou, Jingliang Gu, Canlong Zhang, Huifang Ma

Abstract: Highlights•DyCR-Net can accurately understand satirical meanings in complex satirical scenes.•DyCR-Net dynamically adjust modal correlations for different satirical scenarios.•Unlike traditional fixed satire models, we propose a dynamic adaptive framework.•We validate that DyCR-Net is superior from various aspects and multiple datasets.•DyCR-Net also shows satisfactory results in other similar tasks.