Dynamic interactive multiview memory network for emotion recognition in conversation

Jintao Wen, Dazhi Jiang, Geng Tu, Cheng Liu, Erik Cambria

Abstract: Highlights•This work focuses on the dynamic interactions during information fusion process.•We propose a multiview multimodal fusion that retains more interactive information.•We implement multimodal information fusion within each speaker and between different speakers.•Experiments on IEMOCAP and MELD show state-of-the-art results.