Cold-start item recommendation is a long-standing challenge in recommendation systems. A common remedy is to use a content-based approach, but rich information from raw contents in various forms has not been fully utilized. In this paper, we propose a domain/data-agnostic item representation learning framework for cold-start recommendations, naturally equipped with multimodal alignment among various features by adopting a Transformer-based architecture. Our proposed model is end-to-end trainable completely free from classification labels, not just costly to collect but suboptimal for recommendation-purpose representation learning. From extensive experiments on real-world movie and news recommendation benchmarks, we verify that our approach better preserves fine-grained user taste than state-of-the-art baselines, universally applicable to multiple domains at large scale.
@article{arxiv.2404.13808,
title = {General Item Representation Learning for Cold-start Content Recommendations},
author = {Jooeun Kim and Jinri Kim and Kwangeun Yeo and Eungi Kim and Kyoung-Woon On and Jonghwan Mun and Joonseok Lee},
journal= {arXiv preprint arXiv:2404.13808},
year = {2024}
}