Task-Based Recommendation of Mashup Components

Presentation-oriented mashup applications are usually developed by manual selection and assembly of pre-existent components. The latter are either described on a very technical, functional level, or using informal descriptors, such as tags, which bear certain ambiguities. With regard to the increasing number and complexity of available components, their discovery and integration has become a challenge for non-programmers. Therefore, we present a novel concept for the taskbased recommendation of mashup components, which comprises a more natural, task-driven description of user requirements and a corresponding semantic matching algorithm for universal mashup components. By its realization and integration with an composition platform, we could prove the feasibility and sufficiency of our approach.