VEERA VENKATARAMANI


Research:
I was a graduate student in the Electrical and Computer Engineering Department of The Johns Hopkins University, MD, USA. I worked for The Center for Language and Speech Processing (CLSP).

My advisor was Prof. Bill Byrne. Our field of interest is/was(?) Acoustic Modeling for Speech Recognition.

I am interested in statistical methods for speech recognition in general. I have worked on Speaker Adaptation/Normalization, Pronunciation Modeling and their application to large oral archives .

My thesis is on the application of code-breaking techniques for automatic speech recognition (ASR). Code-breaking is a divide-and-conquer approach to decoding which involves identifying consistent error patterns that a baseline decoder makes. We then train specialized decoders to try and rectify the identified errors. The idea is to intended to locate short-comings of an original speech recognizer and then attempt to rectify them.

We study the use of both Hidden Markov Models (HMMs) and Support Vector Machines (SVMs) as specialized decoders. We have demonstrated the validity of this approach on a small vocabulary task. We have also demonstrated the feasibility of code-breaking for Large Vocabulary Continuous Speech Recognition. We are now looking to obtain further improvements on an LVCSR task.

In summer 2000, I was part of the JHU Workshop. Our group worked on Mandarin Pronunciation Modelling.

Some Publications:
- Veera Venkataramani, and William Byrne, "Lattice Segmentation and Support Vector Machines for Large Vocabulary Continuous Speech Recognition", Accepted, Proc. ICASSP, 2005. [ ps | pdf ]
- Veera Venkataramani, Shantanu Chakrabartty and William Byrne, "GiniSupport Vector Machines For Segmental Minimum Bayes Risk Decoding Of Continuous Speech", Computer Speech and Language, In Press. [ pdf ]
- Veera Venkataramani, Shantanu Chakrabartty and William Byrne, "Support Vector Machines For Segmental Minimum Bayes Risk Decoding Of Continuous Speech", ASRU Workshop, 2003. [ ps | pdf ]
- Veera Venkataramani and William Byrne, "MLLR Adaptation Techniques for Pronunciation Modeling", ASRU Workshop, 2001. [ ps | pdf ]
- William Byrne, Veera Venkataramani, et. al, "Automatic Generation of Pronunciation Lexicons for Mandarin Casual Speech", Proc. ICASSP, 2001. [ ps | pdf ]

I am now employed by the Fair Isaac Corporation as an Analytic Science Lead.


My Resume: resume.pdf


Brief Intro:

From Tamil Nadu, India. No, I do not speak Hindi with any amount of proficiency.

I did my Bachelor's in Electronics and Communication Engineering at REC-Trichy, now NITT.

"There must be more to life than procrastination."


South Indian Cuisine around Baltimore
Contact Info: 

Phone: 410-516-7231(W)
Email at: Veera
 


This page is in memory of Bhavesh, a real fun guy to hang out with; guess the Creator wanted his company also.
Last modified: Mon Jan 1 21:29:27 EST 2007