 |
Center for Nonlinear Studies |
User behavior and interaction data such as search queries, page views, and interactions in social sites, are a potential goldmine of information. Yet, extracting this information is a challenging task, as real user behavior data is complex, varied, and noisy. This talk will address two domains of mining user behavior and interactions: modeling user behavior to improve web search, and modeling interactions and content creation in social and collaborative sites. The first part of the talk will describe machine learning-based techniques for infering relevance and ranking information from web searcher behavior data. The second part will briefly survey our recent work on mining interactions in social media,including estimating content quality, contributor authority, and user satisfaction with the results. BIOSKETCH: Dr. Eugene Agichtein is an Assistant Professor in the Mathematics and Computer Science Department at Emory University. Eugene's research is on information access and retrieval, in particular on understanding and modeling user interactions in web search and social media to improve information access, search, and discovery. Previously, Eugene was a Postdoctoral Researcher in the Text Mining, Search, and Navigation group at Microsoft Research, and a consulting researcher at Yahoo! Research. He has published extensively on information retrieval, web search, and natural language processing. Eugene received a Microsoft "Beyond Search" award, Yahoo! faculty research award, and SIGMOD 2006 Best paper award. Eugene was a senior program committee member (area chair) of SIGIR 2007,2008, and 2009, co-organized the Tutorials at the ACL 2008 conference, and co-chaired the first workshop on Search in Social Media (SSM 2008). Hosted by Johan Bollen (RL) and Marko A. Rodriguez (T5/CNLS)