Printf: Preference Modeling Based on User Reviews with Item Images and Textual Information via Graph Learning

By Hao-Lun Lin et al

Published on June 10, 2023

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Work
3. Printf for Preference Modeling
3.1 Overview
3.2 Cross-Modality Item Modeling (CMIM)
3.3 Review-Aware User Modeling (RAUM)
3.4 User-Item Embedding Propagation and Interaction Modeling (EPIM)
4. Problem Statement
5. Conclusion

Summary

Printf is a novel framework designed to address challenges in preference modeling based on user reviews with item images and textual information. It leverages graph learning techniques to enhance recommender systems by incorporating both textual and visual contents. The framework consists of three main components: Cross-Modality Item Modeling for aligning image and text representations, Review-Aware User Modeling for incorporating user reviews, and User-Item Embedding Propagation and Interaction Modeling for modeling high-order user-item interactions. The proposed framework demonstrates significant improvements in recommendation accuracy based on experiments conducted on publicly available datasets.