Learning Multimodal Word Representation via Dynamic Fusion Methods

TL;DR

This paper introduces three dynamic fusion methods for multimodal word representations, allowing the model to adaptively weight different modalities based on word types, leading to improved semantic understanding.

Contribution

It proposes novel dynamic fusion techniques that assign importance weights to modalities, enhancing multimodal word representations over existing models.

Findings

Proposed methods outperform unimodal baselines.

Proposed methods outperform state-of-the-art multimodal models.

Dynamic weighting improves semantic representation quality.

Abstract

Multimodal models have been proven to outperform text-based models on learning semantic word representations. Almost all previous multimodal models typically treat the representations from different modalities equally. However, it is obvious that information from different modalities contributes differently to the meaning of words. This motivates us to build a multimodal model that can dynamically fuse the semantic representations from different modalities according to different types of words. To that end, we propose three novel dynamic fusion methods to assign importance weights to each modality, in which weights are learned under the weak supervision of word association pairs. The extensive experiments have demonstrated that the proposed methods outperform strong unimodal baselines and state-of-the-art multimodal models.