Modeling Intention in Email: Speech Acts, Information Leaks and User Ranking Methods

February 28, 2008
Vitor Carvalho | Carnegie Mellon University

Email management has a fundamental role in work productivity. In this talk I will present evidence that machine learning techniques can be effectively used to improve email management by modeling different aspects of user intention. First I’ll propose a taxonomy of Speech Acts applied to email communication, or “email acts”, and show that email act classification can be largely automated, potentially leading to better email prioritization and management. Then I’ll describe how machine learning can be used to reduce the chances of high-cost email addressing errors. One type of high-cost error is an “email leak” , i.e., mistakenly sending a sensitive email message to the wrong recipient. Another type of addressing error is forgetting to add an important collaborator as recipient, a likely source of costly misunderstandings and communication delays. I’ll present different approaches to these problems, including solutions based on rank aggregation and new (re)ranking techniques.

Speaker Details

Vitor R. Carvalho is a PhD candidate in the School of Computer Science at Carnegie Mellon University. Vitor’s research interests include text and data mining, machine learning, information extraction, information retrieval and email management. He has authored several papers on these topics, and served on the program committee of WWW, AAAI, ACL and CEAS. From 2000 to 2003 he worked at the Research and Development Center of Ericsson Telecom in Sao Paulo, Brazil. Vitor is currently organizing the Workshop on Enhanced Messaging on AAAI-2008.