GiveALink: a social
bookmarking site where people can donate their bookmarks
to the Web community and to science
(More
information about the project)
My current work focuses on providing tools that make use of the implicit relationships found in social-tagging systems to help web users bookmark, navigate, and explore the web. More specifically, this involves the recommendation of tags for an individual resource (bookmark), recommendation of resources potentially interesting to the individual user, visualization of relationships extracted from bookmarks and tags, and design of interfaces that allow users to understand and make use of the available information. Such tools can provide benefits in both directions: users can manage their tagged bookmarks and find resources of interest, encouraging participation which provides more data to GiveALink on which to base the recommendations and relationship networks.
During the summer of 2009, I demonstrated a Firefox extension I have been developing for GiveALink (with help and input from Ciro Cattuto and Wouter Van den Broeck) and a tagging game (in development) at both the Hypertext and SIGIR conferences. The Firefox extension is available for download and use. With it, users can manage their GiveALink bookmarks, search their GiveALink bookmarks, search GiveALink, and navigate the web using the relationships extracted from the bookmarks of GiveALink users as additional links.
During the summer of 2008, the GiveALink group (including now-alum Benjamin Markines) presented a paper on our work on making the similarity network building scalable, by making updates to the network incremental, and on testing different such measures at HT 2008. (The resulting paper was nominated for best paper.) An additional short paper, co-authored with Justin Donaldson and Michael Conover, explored visualizing similarity relationships among search results to assist in exploratory search.
I also previously collaborated with Ana Maguitman, Fil Menczer, and Alessandro Vespignani on a way to automatically measure similarity between two urls classified in an ontology (such as the Open Directory Project) for use in automatically testing new similarity measures based on the content and/or link structure of the pages. This work led to a paper which was a finalist for the best paper award at WWW2005.