Self-Supervised Multimodal Opinion Summarization

TL;DR

This paper introduces MultimodalSum, a self-supervised framework that leverages both text and non-text review data, including images and metadata, to generate more comprehensive opinion summaries.

Contribution

It proposes a novel multimodal training pipeline with separate encoders for each modality and end-to-end fusion, enhancing opinion summarization with non-text data.

Findings

MultimodalSum outperforms text-only models on Yelp and Amazon datasets.

Pretraining on individual modalities improves overall summarization quality.

Incorporating non-text data significantly enhances summary informativeness.

Abstract

Recently, opinion summarization, which is the generation of a summary from multiple reviews, has been conducted in a self-supervised manner by considering a sampled review as a pseudo summary. However, non-text data such as image and metadata related to reviews have been considered less often. To use the abundant information contained in non-text data, we propose a self-supervised multimodal opinion summarization framework called MultimodalSum. Our framework obtains a representation of each modality using a separate encoder for each modality, and the text decoder generates a summary. To resolve the inherent heterogeneity of multimodal data, we propose a multimodal training pipeline. We first pretrain the text encoder--decoder based solely on text modality data. Subsequently, we pretrain the non-text modality encoders by considering the pretrained text decoder as a pivot for the homogeneous representation of multimodal data. Finally, to fuse multimodal representations, we train the entire framework in an end-to-end manner. We demonstrate the superiority of MultimodalSum by conducting experiments on Yelp and Amazon datasets.