Google is supporting UTEP to research Vocal Indications of Feelings and Stance in Spoken Dialog.
Google Ads helps customers connect with businesses, and in some cases they are interested in statistics on whether the outcomes are good or bad. When telephone conversations are involved, they can be a source of information, but Google needs methods that are automatic, without involving humans listening in. Professor Ward's group has previously had success in automatically detecting dialog activities and stances for NSF and Darpa projects. In this project they will collect data and model the non-lexical, prosodic aspects of speech to address this problem.