|
2:00 p.m.
2168 A.V. Williams Building
For More Information:
Carol Espy-Wilson
301 405 7411
espy@umd.edu
Analytics for Contact Centers
Ashish Verma
Research Staff Member
IBM Research India
Host
Carol Espy-Wilson
Abstract
In this talk, we would cover some of the research challenges involved in contact center analytics. In the first half of the talk, I would present methods to evaluate spoken English skills of a person, in particular, spoken fluency and syllable stress pattern. The proposed fluency evaluation technique (ICASSP 2009) uses a combination of novel prosodic and lexical features to produce performance comparable to human assessors in contact centers. For syllable stress evaluation, we will see how using a categorization based on acoustic-phonetics paradigm improves syllable classification accuracy (Speech Communication 2009). In the second part of the talk, I would present the work on how to utilize a human refined set of text clusters to perform improved clustering in a related domain (ICDM 2009). I would also highlight the business importance of these techniques in the contact center domain.
Biography
Ashish Verma received his Ph. D degree from Indian Institute of Technology, Delhi in 2006 in the area of voice conversion. He is working as a Research Staff Member at IBM Research - India, since 1998 and currently managing the Contact Center and Speech Technologies group. At IBM, he has been involved in several projects, such as automatic spoken language evaluation, audio-visual speech recognition, voice of customer analytics, speech recognition system for Indian languages, etc. His areas of interest include speech processing, text mining and pattern recognition. Ashish is a senior member of IEEE.
This Event is For: Clark School • Graduate • Faculty • Post-Docs • Alumni
|
