A Dataflow-Pattern-Based Recommendation Approach for Data Service Mashups

Abstract

Though there are some existing data service mashup tools, it is still challenging for those developers with no or little programming skills to develop data service mashups to solve the situational and ad-hoc business problems. This paper focuses on the problem of interactively recommending useful assistance at every step during the development of data service mashups under the condition that the mashup plan can't be determined in advance. This paper analyzes the problem with a motivating scenario, introduces the core definitions and an approach to dataflow pattern based recommendation. Inspired by the idea that there exist dataflow patterns for certain integration functionalities, several types of data service mashup patterns are defined. Then the interactively data service mashup recommendation method is proposed based on them. We also associate a set of tags to represent the situation and the inputs and outputs of data service model, and incorporate it in the recommendation method. Experiment results show that the dataflow pattern based recommendation approach for data service mashup is effective.