A tremendous amount of semi-structured data is available today on the web but is not necessarily in a form which is suitable for a user's tasks. For example, a website may show a listing of local events but a user wants to filter out those which are too far from him. To address this problem, we are developing a tool called Marmite that helps users extract data from web pages and create new applications using a dataflow architecture in a manner similar to Unix pipes. In this paper, we describe formative user studies, some evaluations of low-fidelity prototypes, and a set of design recommendations for this tool. Keywords End-user programming, web automation, web browsers ACM Classification Keywords H5.2 [Information interfaces and presentation]: User Interfaces. - Graphical user interfaces; D.2.6 [Programming Environments]: Interactive environments;
Jason I. Hong, Jeffrey Wong