|
The more than one-trillion pages on the web today provide a corpus of design examples unparalleled in human history. Current tools for sampling from these existing designs require manually manipulating the HTML source. To enable novices and experts alike to more creatively use examples, this paper introduces higher-level techniques for incorporating existing content into new web designs. We contribute a machine learning algorithm that computes semantic correspondences between the visual layouts of web pages. These correspondences can subsequently guide the transfer of design between pages. The features the algorithm uses are derived from HTML properties, vision-based analysis, and the hierarchical structure encoded in the Document Object Model. The algorithm trains on a set of manually-specified page mappings. With this algorithm, any HTML can be a design source. This design-borrowing approach is manifest in Bricolage, a tool enabling users to integrate extant design elements into their own work. |
|
|
Juho Kim, Ranjitha Kumar, Scott R. Klemmer. Crowdsourcing Interface for Collecting Correspondences of Web Pages. Poster at UIST'09, Victoria, BC, Canada, 2009. Ranjitha Kumar, Juho Kim, Scott R. Klemmer. Automatic Ratargeting of Web Page Content. In Extended Abstracts of CHI2009, Boston, MA, 2009. |
|