RBox: An experimentation tool for creating event-driven recommender algorithms for web 2.0

Abstract

This article introduces a software environment called RBox, built to experiment with recommender systems (RS), regardless of the application domain. In spite of the ubiquity of RS on the Web 2.0 this research field still lacks a unique way of representing collective intelligence. To solve this problem, this article adopts a generic event-driven approach providing a unique RBox data schema. Thus, it is possible to achieve the abstraction of collaborative events that occur on Web 2.0 such as ranking, tagging and voting. A comparison with other tools illustrates the contribution of RBox to the RS field. For instance, this tool enables reusing algorithms and executing experiments that were originally intended for a specific application domain, for other ones. Finally, considering RS tools’ limitations, the next versions of RBox will integrate ubiquitous computing and context-aware recommender systems.