Efficient, Interactive Recommendation of Mashup Composition Knowledge

In this paper, we approach the problem of interactively querying and recommending composition knowledge in the form of re-usable composition patterns. The goal is that of aiding developers in their composition task. We specifically focus on mashups and browser-based modeling tools, a domain that increasingly targets also people without profound programming experience. The problem is generally complex, in that we may need to match possibly complex patterns on-the-fly and in an approximate fashion. We describe an architecture and a pattern knowledge base that are distributed over client and server and a set of client-side search algorithms for the retrieval of step-by-step recommendations. The performance evaluation of our prototype implementation demonstrates that - if sensibly structured - even complex recommendations can be efficiently computed inside the client browser. © 2011 Springer-Verlag