Generic system architecture for context-aware, distributed recommendation

Abstract

In the existing literature on recommender systems, it is difficult to find an architecture for large-scale implementation. Often, the architectures proposed in papers are specific to an algorithm implementation or a domain. Thus, there is no clear architectural starting point for a new recommender system. This paper presents an architecture blueprint for a context-aware recommender system that provides scalability, availability, and security for its users. The architecture also contributes the dynamic ability to switch between single-device (offline), client-server (online), and fully distributed implementations. From this blueprint, a new recommender system could be built with minimal design and implementation effort regardless of the application.