The document describes a social information streaming and recommendation system. It discusses components like URL sources, topic relevance scores, social network scores, and a recommendation engine. It analyzes different design choices for each component and evaluates their effects on recommendation accuracy through a live deployment. Key findings are that combining followee-of-followees URLs, self-topic relevance scores, and social voting leads to the most accurate recommendations. The system demonstrates that algorithmic choices impact not only accuracy but also user experiences.
1 of 27
Downloaded 28 times
More Related Content
Zerozero88 Twitter URL Item Recommender
1. +University of Palo Alto Research Center! ^MIT CSAIL!
Minnesota!
8/23/10 CHI2010 ZeroZero88.com 1
2. ?? Social
Information
Streams!
8/23/10 CHI2010 ZeroZero88.com 2
27. Acknowledgements!
?? ASC folks at PARC!
?? All users of Zerozero88.com!
?? My labmates at the University for helping me
prepare the presentation!
?? ¡ and reach me at jilin@cs.umn.edu for
questions!!