Description
This paper investigates architectural properties required for supporting automatic service composition. First, composable service architecture will be described, based on modeling Webn Services as abstract machines supported by formally defined composition operators. Based on the proposed infrastructure, we give several options for achieving automatic service composition by treating it as a search problem. Namely, basic heuristic, probabilistic, learning-based, decomposition and bidirectional automatic composition mechanisms will be presented and compared. Finally, it discusses the impact and outlook for automatic composition.