Solving cold start in news recommendations: a RippleNet-based system for large scale media outlet

TL;DR

This paper introduces a scalable RippleNet-based recommender system for large media platforms that effectively addresses cold-start issues by integrating content-based embeddings and deploying on cloud infrastructure.

Contribution

It presents a novel integration of content embeddings into RippleNet for cold-start mitigation in large-scale media recommendation systems.

Findings

Effective cold-start handling for new content.

Scalable system architecture using cloud services.

Improved recommendation quality for unseen items.

Abstract

We present a scalable recommender system implementation based on RippleNet, tailored for the media domain with a production deployment in Onet.pl, one of Poland's largest online media platforms. Our solution addresses the cold-start problem for newly published content by integrating content-based item embeddings into the knowledge propagation mechanism of RippleNet, enabling effective scoring of previously unseen items. The system architecture leverages Amazon SageMaker for distributed training and inference, and Apache Airflow for orchestrating data pipelines and model retraining workflows. To ensure high-quality training data, we constructed a comprehensive golden dataset consisting of user and item features and a separate interaction table, all enabling flexible extensions and integration of new signals.