Holistic Multi-modal Memory Network for Movie Question Answering

TL;DR

This paper introduces the Holistic Multi-modal Memory Network (HMMN), a novel framework that fully integrates multi-modal context, questions, and answer choices for improved movie question answering accuracy.

Contribution

The paper proposes a comprehensive multi-hop attention framework that considers all data sources simultaneously, enhancing multi-modal reasoning in question answering tasks.

Findings

Achieves state-of-the-art accuracy on MovieQA dataset.

Demonstrates the effectiveness of holistic reasoning and attention strategies.

Shows significant improvements over partial interaction models.

Abstract

Answering questions according to multi-modal context is a challenging problem as it requires a deep integration of different data sources. Existing approaches only employ partial interactions among data sources in one attention hop. In this paper, we present the Holistic Multi-modal Memory Network (HMMN) framework which fully considers the interactions between different input sources (multi-modal context, question) in each hop. In addition, it takes answer choices into consideration during the context retrieval stage. Therefore, the proposed framework effectively integrates multi-modal context, question, and answer information, which leads to more informative context retrieved for question answering. Our HMMN framework achieves state-of-the-art accuracy on MovieQA dataset. Extensive ablation studies show the importance of holistic reasoning and contributions of different attention strategies.