Learning Question-Guided Video Representation for Multi-Turn Video Question Answering

TL;DR

This paper introduces a question-guided video representation method for multi-turn video question answering, which efficiently summarizes videos based on questions and improves answer accuracy.

Contribution

It proposes a novel question-guided video encoding module that enhances multi-turn video question answering by focusing on relevant video segments.

Findings

Achieves state-of-the-art results on AVSD dataset.

Outperforms existing models in automatic evaluation metrics.

Efficiently encodes video segments relevant to questions.

Abstract

Understanding and conversing about dynamic scenes is one of the key capabilities of AI agents that navigate the environment and convey useful information to humans. Video question answering is a specific scenario of such AI-human interaction where an agent generates a natural language response to a question regarding the video of a dynamic scene. Incorporating features from multiple modalities, which often provide supplementary information, is one of the challenging aspects of video question answering. Furthermore, a question often concerns only a small segment of the video, hence encoding the entire video sequence using a recurrent neural network is not computationally efficient. Our proposed question-guided video representation module efficiently generates the token-level video summary guided by each word in the question. The learned representations are then fused with the question to generate the answer. Through empirical evaluation on the Audio Visual Scene-aware Dialog (AVSD) dataset, our proposed models in single-turn and multi-turn question answering achieve state-of-the-art performance on several automatic natural language generation evaluation metrics.