Pre-training Graph Transformer with Multimodal Side Information for Recommendation

TL;DR

This paper introduces a pre-training method for graph transformers that leverages multimodal item side information and user activity-based relations to improve recommendation accuracy and other downstream tasks.

Contribution

It proposes a novel pre-training strategy using a homogeneous item graph with multimodal data and a new sampling algorithm, MCNSampling, for enhanced item representation learning.

Findings

PMGT outperforms baseline models in recommendation accuracy

Effective utilization of multimodal side information improves downstream task performance

Case study confirms scalability and real-world applicability

Abstract

Side information of items, e.g., images and text description, has shown to be effective in contributing to accurate recommendations. Inspired by the recent success of pre-training models on natural language and images, we propose a pre-training strategy to learn item representations by considering both item side information and their relationships. We relate items by common user activities, e.g., co-purchase, and construct a homogeneous item graph. This graph provides a unified view of item relations and their associated side information in multimodality. We develop a novel sampling algorithm named MCNSampling to select contextual neighbors for each item. The proposed Pre-trained Multimodal Graph Transformer (PMGT) learns item representations with two objectives: 1) graph structure reconstruction, and 2) masked node feature reconstruction. Experimental results on real datasets demonstrate that the proposed PMGT model effectively exploits the multimodality side information to achieve better accuracies in downstream tasks including item recommendation, item classification, and click-through ratio prediction. We also report a case study of testing the proposed PMGT model in an online setting with 600 thousand users.