Hybrid techniques for web APIs recommendation

• Authors: Michele Melchiori
• Venue: Proceedings of the 1st International Workshop on Linked Web Data Management
• Year: 2011

Currently on the Web there is a proliferation of software components with public APIs that make available ready-to-use functionalities and access to contents. These APIs can be exploited to build applications, called mashups, that are gaining a lot of interest as an opportunity to integrate contents and application logics from independent sources in short-living situational applications. Although existing mashup frameworks, like Google's Mashup Editor or Yahoo Pipes, provide valid support for development of new applications from given content sources, designers often have to deal with large component repositories in which APIs for components are usually featured by semantic and technological heterogeneities. This situation creates disorientation in a designer that has to select manually components and finding the better way to compose them in a mashup. As a contribution to solve this problem, we introduce hybrid techniques that combine semantics and collective knowledge to organize component descriptors in a registry according to coupling links. Then, we discuss how this organization can be exploited to implement proactive component recommendation for the design of mashup applications.