Knowledge-Aware Causal Inference Network for Visual Dialog

What is it about?

We propose a novel framework named Knowledge-Aware Causal Inference Network (KACI-Net) for solving the complex scenarios in visual dialog. From causality view, we prove the necessity of commonsense knowledge and design the coreference resolution module for constructing a causal relation among variables.

Featured Image

Photo by Volodymyr Hryshchenko on Unsplash

Photo by Volodymyr Hryshchenko on Unsplash

Why is it important?

We propose a Knowledge-Aware Causal Inference Network (KACI-Net) by incorporating the commonsense knowledge for improving the reasoning accuracy in complex scenarios. We analyze the effect of commonsense knowledge which is extracted from history dialog.