• Home
• Attending
• Authors
• Sponsors
• Recruiters
• Exhibitors
• Organizers
• Press

Papers/Notes: Machine Learning and Web Interactions

Tuesday, April 13
4:30 PM - 6:00 PM

Interactive Optimization for Steering Machine Classification
Ashish Kapoor, Microsoft Research, USA
Bongshin Lee, Microsoft Research, USA
Desney Tan, Microsoft Research, USA
Eric Horvitz, Microsoft Research, USA

ManiMatrix is an interactive system that allows interactive refinement of classification boundaries in a multiclass setting. The system interweaves visualization, interaction, and optimization to steer classification according to users preferences.

 A Longitudinal Study of How Highlighting Web Content Change Affects People's Web Interactions
Jaime Teevan, Microsoft Research, USA
Susan T. Dumais, Microsoft Research, USA
Daniel J. Liebling, Microsoft Research, USA

Longitudinal study shows that highlighting changes in Web content leads to increased Web page revisitation, and improved perception and use of content change within the revisited content.

Examining Multiple Potential Models in End-User Interactive Concept Learning
Saleema Amershi, University of Washington, USA
James Fogarty, University of Washington, USA
Ashish Kapoor, Microsoft Research, USA
Desney Tan, Microsoft Research, USA

Re-examines a traditional interactive machine learning focus on ?what class is this object??, broadening interaction to include examining multiple potential models. This approach improves the quality of end-user trained models.

Signed networks in social media
Jure Leskovec, Stanford University, USA
Daniel Huttenlocher, Cornell University, USA
Jon Kleinberg, Cornell University, USA

We analyze on-line social networks where links can be either positive or negative. We extend theories from social psychology to explore implications of these signed networks for social computing applications.