Towards communities of practice for mashups
Proceedings of the 3rd and 4th International Workshop on Web APIs and Services Mashups
Building Mashups by Demonstration
ACM Transactions on the Web (TWEB)
| Hi-index | 0.00 |
Accurately integrating the information available on the Internet can provide valuable insights useful in decision making. However, the information one needs is usually scattered among multiple web sites. It can be time-consuming to access, clean, combine, and make sense of that data manually. The latest generation of WWW tools and services enables web users to generate web applications that combine content from multiple sources and provide them as unique services that suit their individual needs. This type of web applications is referred to as a Mashup. To create Mashups, integration systems must be able to walk users through five separate problems: data extraction, source modeling, data cleaning, data integration, and data display. While there exist attempts to facilitate the process of building information integration applications, none is sufficiently easy to use to enable a web user to build an end-to-end information integration application. As a result, a casual user is put off by the time, effort, and expertise needed to build a Mashup. In this thesis, I make three contributions that address the problem of building Mashups so that a casual user can easily build one efficiently without having to know any programming language. The first is a consolidated approach that uses a database to link Mashup building problems together. Under this approach, the solution from solving a problem in one area will be leveraged to solve a problem in the next area. The second is a table paradigm for building Mashups, where a user incrementally builds a Mashup by filling table cells with examples instead of specifying programming operations. The third is a query formulation technique that uses constraints to help users build complicated queries by specifying examples. To validate the approach in this thesis, I have done an extensive evaluation involving more than twenty subjects. The evaluation compared the performace of Karma, the system implementation of my approach, against current state-of-the-art systems: Dapper and Pipes. The result shows that Karma is at least three times faster than Dapper/Pipes on three representative Mashup building tasks. In addition, Karma also allows users to build Mashups that they fail to build using the other systems.