Recommending from Actions, Not Content: New Approaches in Social Recommendation
by David Ayman Shamma
Internet Experiences group
Content recommendation engines to date are informed by explicit annotations, such as ratings and comments, derived from social actions. Often, the semantics of social recommendation turns to examining tag clusters, co-occurrence, and other aggregate social features. Despite the success of these collaborative-filtering techniques, they are a weak proxy of the deep social interaction that accompanies content sharing. Newer social media systems, especially those that encourage real-time interactions, can provide us with better fidelity data on which to make content recommendations. My recent investigations in using the metadata from real-time interactions have produced highly accurate predictions about the nature and characteristics of the shared content. These accurate predictions can thus be extended to provide better recommendations for content. In this talk, I will present new methods of recommendation based on the social signals derived from conversational sharing of content. I will illustrate these methods through the examination of YouTube videos that are shared in a synchronous environment over Instant Messaging; recommendations are delivered based on the implicit communicative interactions that happens around media and not the media's content itself. Better understanding of these deeper and more connected sharing constructs has great significance and implications for the development of future recommendation systems and interaction experiences.
Speaker's short bio
David Ayman Shamma is a research scientist in the Internet Experiences group at Yahoo! Research. He researches synchronous environments and connected experiences both online and in-the-world. Focusing on creative expression and sharing frameworks, he designs and prototypes systems for multimedia-mediated communication, as well as, develops targeted methods and metrics for understanding how people communicate online in small environments and at web scale. Ayman is the creator and lead investigator on the Yahoo! Zync project. Outside of the lab, Ayman has created several interactive art installations that have been reviewed internationally by The New York Times, International Herald Tribune, and Chicago Magazine.
Ayman holds a B.S./M.S. from the Institute for Human and Machine Cognition at The University of West Florida and a Ph.D. in Computer Science from the Intelligent Information Laboratory at Northwestern University. Before Yahoo!, he was an instructor at the Medill School of Journalism; he has also taught courses in Computer Science and Studio Art departments. Prior to receiving his Ph.D., he was a visiting research scientist for the Center for Mars Exploration at NASA Ames Research Center.
Building Social Recommenders for Delicious and Twitter
by Ed H. Chi
Area Manager and Principal Scientist
Augmented Social Cognition Area
Palo Alto Research Center, USA
Information search and discovery engines now rely on not just personalized models of interests, but also the social cues created by a large number of people. The attention traces left behind by people are valuable navigational signposts for building social recommenders. We can take advantage of the fact that these traces are being generated in a social context, with networks of friends and friends-of-friends as potential audiences and transceivers. In this talk, I will talk about the use of these cues in two systems:
First, in MrTaggy.com, we used the social cues from social bookmarks sites. Social tagging arose out of the need to organize found information that is worth revisiting. The collective behavior of users who tagged contents offer a good basis for recommendation engines. We used information theory and probabilistic graph models to pre-compute recommendations, and evaluated this exploratory browsing system in the lab using end-user learning metrics.
Second, in Zerozero88.com, we constructed a tweet recommender for Twitter users. In a modular approach, we explored three separate dimensions in designing such a recommender: content sources, topic interest models for users, and social voting. We evaluated the system by having twitter users rank the recommendations we gave them over a 3 week period. The results show how recommenders can profitably integrate social cues.
Speaker's short bio
Ed H. Chi is a Principal Scientist at the (Xerox) Palo Alto Research Center, where he is also the Area Manager for the Augmented Social Cognition Area. He leads the group in understanding how Web2.0 and Social Computing systems help groups of people to remember, think and reason. For example, the group has studied the underlying mechanisms in online social systems such as Wikipedia, Twitter, and social tagging sites. Ed completed his three degrees (B.S., M.S., and Ph.D. summa cum laude) in 6.5 years from University of Minnesota, and has been doing research on user interface software systems since 1993.
He has also worked on information visualization, computational molecular biology, ubicomp, and recommendation/search engines. With over 20 patents and 80 research articles, he has been featured and quoted in the press, including the Economist, Time Magazine, LA Times, and the Associated Press. In his spare time, Ed is an avid Taekwondo martial artist, photographer, and snowboarder.