Monday November 17 4:00 - 4:50 PM Kelley 1001
Izhak (Zak) Shafran Assistant Professor Center for Spoken Language Understanding Oregon Health and Science University Beyond Words: Recognizing Affect (or Emotion) from Speakers' Voice To design more natural spoken language interface, machines should be able to detect affect, an innate and universal human ability. Affect recognition requires tackling three challenges: (a) an appropriate similarity metric to measure distance between utterances, (b) a representation that captures inherent ambiguities in speech, and (c) mechanism for fusing different modalities (e.g. what is said and how it is said). This talk will describe a framework that address the above three challenges for recognizing affect, and more generally, for any speaker characteristics. Experimental results will be reported on "How May I Help You", a call-center application that catered to two millions of customers per month during its deployment. Biography: Izhak (Zak) Shafran is currently an Assistant Professor at Oregon Health & Science University and is a member of the Center for Spoken Language Understanding. His research is largely focused on modeling speech, specifically in the context of large vocabulary speech-to-text, acoustic modeling, spoken term detection, prosody modeling, and language recognition. The application areas of his research ranges from extracting information from speech to functional assessment for medical applications. After obtaining his doctoral degree from University of Washington, he joined AT&T Labs Research in the Speech Algorithms Group. During his tenure at AT&T, he spent a summer at LIMSI (France) as a visiting professor at the University of Paris-South. Subsequently, he was a research faculty in the Center for Language and Speech Processing in the Johns Hopkins University, before joining OHSU.
_______________________________________________ Colloquium mailing list [email protected] https://secure.engr.oregonstate.edu/mailman/listinfo/colloquium
