MARec: Metadata Alignment for cold-start Recommendation

TL;DR

This paper introduces Metadata Alignment, a simple yet effective method for cold-start recommendation that leverages content metadata to significantly improve ranking performance, outperforming state-of-the-art methods on multiple datasets.

Contribution

The paper presents a novel framework that enhances cold-start recommendation by integrating metadata, achieving substantial improvements over existing methods and maintaining competitiveness in warm scenarios.

Findings

Outperforms SOTA on 4 cold-start datasets with +8.4% to +53.8% gains.

Semantic features provide an additional +46.8% to +105.5% improvement.

Highly competitive in warm set-ups, with only 0.8% average difference from SOTA.

Abstract

For many recommender systems, the primary data source is a historical record of user clicks. The associated click matrix is often very sparse, as the number of users x products can be far larger than the number of clicks. Such sparsity is accentuated in cold-start settings, which makes the efficient use of metadata information of paramount importance. In this work, we propose a simple approach to address cold-start recommendations by leveraging content metadata, Metadata Alignment for cold-start Recommendation. We show that this approach can readily augment existing matrix factorization and autoencoder approaches, enabling a smooth transition to top performing algorithms in warmer set-ups. Our experimental results indicate three separate contributions: first, we show that our proposed framework largely beats SOTA results on 4 cold-start datasets with different sparsity and scale characteristics, with gains ranging from +8.4% to +53.8% on reported ranking metrics; second, we provide an ablation study on the utility of semantic features, and proves the additional gain obtained by leveraging such features ranges between +46.8% and +105.5%; and third, our approach is by construction highly competitive in warm set-ups, and we propose a closed-form solution outperformed by SOTA results by only 0.8% on average.