Specializing Joint Representations for the task of Product Recommendation

TL;DR

This paper introduces a modular approach to create unified product embeddings by fusing content and collaborative signals, improving retrieval-based recommendations especially in cold-start and cross-category scenarios.

Contribution

It presents a novel late-fusion method for combining modality-specific embeddings into a joint representation, maintaining modularity for practical deployment.

Findings

Effective in cold-start scenarios

Performs well on cross-category recommendations

Achieves strong results on large shopping dataset

Abstract

We propose a unified product embedded representation that is optimized for the task of retrieval-based product recommendation. To this end, we introduce a new way to fuse modality-specific product embeddings into a joint product embedding, in order to leverage both product content information, such as textual descriptions and images, and product collaborative filtering signal. By introducing the fusion step at the very end of our architecture, we are able to train each modality separately, allowing us to keep a modular architecture that is preferable in real-world recommendation deployments. We analyze our performance on normal and hard recommendation setups such as cold-start and cross-category recommendations and achieve good performance on a large product shopping dataset.