SRS – Explainable and Privacy-Preserving Smart Recommendation System for Digital Publishers

A production-oriented AI recommendation platform for digital publishers and content creators, developed by SQUARE1 within EARS. The solution analyses heterogeneous and unstructured information from digital media, news, social platforms and user interactions to identify topics, trends, audience sentiment and content preferences. It combines NLP, Knowledge Graphs, hybrid recommendation models and Federated Learning to generate personalised and explainable recommendations for content creation strategies, while preserving user privacy and supporting GDPR compliance. The result can be integrated with SQUARE1's Publisher Plus CMS or consumed by third-party platforms through interoperable APIs.

The exploitable result integrates three main technological blocks: AI Smart User Data Consumer, Intelligent Data Processor and Smart Recommendation System. The platform captures and processes unstructured information, extracts entities, topics, sentiment and semantic relationships, and represents this knowledge through structured data and Knowledge Graphs. Hybrid recommendation algorithms combine content-based and collaborative approaches, while Federated Learning enables collaborative model training without requiring centralisation of sensitive raw data. The recommendation layer generates personalised outputs together with explainability information, allowing content creators to understand the factors behind a recommendation. APIs and integration services enable external applications and CMS platforms to consume the resulting recommendation services.

Integration follows an API-based and interoperable architecture. External systems must provide compatible content, user-interaction or contextual information through the defined service interfaces and respect the corresponding data models and authentication mechanisms. Deployment requires compliance with SQUARE1/EARS security and privacy requirements, particularly user consent, data minimisation and GDPR-related processing rules. The architecture is designed to support integration with Publisher Plus as well as third-party CMSs and recommendation environments; interoperability and external service integration were specifically addressed within WP5.

Primary customers are digital publishers, media organisations, online newspapers, sports media platforms, content aggregators and organisations operating high-traffic digital content platforms. The principal users are editors, journalists, content strategists and content creators seeking data-driven guidance about topics, audience interests and content creation strategies. The solution is particularly suitable for existing Publisher Plus customers but, due to its interoperable architecture, can also be commercialised independently of SQUARE1's CMS.

The result is intended for commercial exploitation by SQUARE1, either as an integrated service within Publisher Plus or through integration with external digital platforms. The commercial model foreseen in the project is based on licensing the Smart Recommendation System (SRS) together with the necessary front-end/back-end integration services. For existing Publisher Plus customers, the recommendation functionality may be provided as an additional plug-in/service with limited adaptation effort. Reuse by external platforms requires compliance with the API specifications, security/privacy requirements and applicable software/IP licensing conditions.