October 31, 2013
Design Mining the Web
The billions of pages on the Web today provide an opportunity to understand design practice on a truly massive scale: each page comprises a concrete example of visual problem solving, creativity, and aesthetics. In recent years, data mining and knowledge discovery have revolutionized the Web, driving search engines and recommender systems that are used by millions of people every day. However, data mining traditionally focuses on the content of Web pages, ignoring how that content is presented. What can we learn from miningdesign?
This talk presents design mining for the Web, and presents a scalable platform for Web design mining called Webzeitgeist. Webzeitgeist consists of a repository of pages processed into data structures that facilitate large-scale design knowledge extraction. With Webzeitgeist, users can find, understand, and leverage visual design data in Web applications. I’ll demonstrate how software tools built on top of Webzeitgeist can be used to dynamically curate design galleries, search for design alternatives, retarget content between page designs, and even predict the semantic role of page elements from design data. As more and more creative work is done digitally and shared in the cloud, Webzeitgeist provides a concrete illustration of how design mining principles can be applied to benefit content creators and consumers. To learn more, visit webzeitgeist.stanford.edu.
Ranjitha Kumar is the Chief Scientist at Apropose, Inc., a Bay Area startup she co-founded to build data-driven design software for the Web. She holds BS and PhD degrees from Stanford University, and will begin as an Assistant Professor in the Department of Computer Science at the University of Illinois at Urbana-Champaign in fall 2014. Her work has received best paper awards or nominations at both of the premier HCI conferences (CHI and UIST), and been recognized by the machine learning community through invited papers at IJCAI and ICML.