VAIBHAVA GOEL Center for Language and Speech Processing Johns Hopkins University 320 Barton Hall, 3400 N. Charles Street Baltimore, MD 21218 Email: vgoel@jhu.edu, vgoel@bme.jhu.edu Cell Phone: (410) 905-1877 Lab Phone: (410) 516-5458,5409 Lab Fax: (410) 516-5050 Home phone: (410) 420-1352 RESEARCH INTERESTS Automatic Speech Recognition; Statistical Modeling and Learning; Classification and Decision Theory; Information Theory and Statistics; EDUCATION Ph.D., July 2000 (expected), Biomedical Engineering, Johns Hopkins University Dissertation: Minimum Bayes-risk automatic speech recognition. Advisors: Dr. William J. Byrne and Dr. Frederick Jelinek. GPA: 4.0/4.0 M.S.E., May 1995, Biomedical Engineering, Johns Hopkins University Thesis: A novel technique for EEG signal processing. Advisor: Dr. Nitish V. Thakor. GPA: 3.95/4.0 B.Tech., May 1993, Electrical Engineering, Indian Institute of Technology, Kanpur, India Thesis: Automated strain analysis using techniques of image processing. Advisor: Dr. S. K. Mullick. GPA: 10.0/10.0 WORK EXPERIENCE The Bell Laboratories, Lucent Technologies, Murray Hill, NJ ----------------------------------------------------------- Summer Intern, May 1997 - Jul 1997 Project: Maximum likelihood lexical modeling. Supervisor: Dr. Frank K. Soong Summer Intern, Jun 1996 - Aug 1996 Project: Single pass real time A-star decoder implementation. Supervisor: Dr. Frank K. Soong Indian Telephone Industries, Naini, India ----------------------------------------- Summer Engineer, May 1992 - Jul 1992 Project: Optimization of production and inventory control of MARR systems. TEACHING & RESEARCH Center for Language and Speech Processing, Johns Hopkins University ------------------------------------------------------------------- Research Assistant, Jan 1996 - Present Design of task specific speech recognizers under the minimum Bayes-risk (MBR) classification framework. Formulated a novel prefix-tree based multi-stack A-star algorithm to implement the MBR decoders. Shown that these decoders yield significant error rate reduction on tasks such as minimum word error rate transcription, keyword spotting, named entity detection, and gene identification from genomic DNA. Developed the idea of segmental MBR decoding in which a complex MBR decoder is approximated by a sequence of simple MBR decoders. Demonstrated that the well known voting techniques of ROVER and voting on lattices are instances of segmental MBR decoding, and developed their extensions. Researcher, Jul 1997 - Aug 1997 Workshop on Innovative Techniques for LVCSR; Syllable based speech recognition. Teaching Assistant, Feb 1997 - May 1997 & Feb 1996 - May 1996 Neuroengineering : A course on application of engineering ideas for diagnosis and treatment of neuropathologies. Biomedical Instrumentation Laboratory, Johns Hopkins University --------------------------------------------------------------- Research Assistant, Sep 1993 - Aug 1995 LPC analysis and neuronal correlates of anomalies in electroencephalograms. Involved applied signal processing, instrumentation, and neuronal simulations and modeling. Teaching Assistant, Feb 1995 - May 1995 & Sep 1994 - Dec 1994 Biomedical Instrumentation : A course on instrumentation for medical applications. HONORS & AWARDS Abel Wolman Fellowship, Johns Hopkins University (95-96). Tuition fellowship, Johns Hopkins University (93-95). Best outgoing undergraduate student in electrical engineering at Indian Institute of Technology Kanpur (IITK), 1993. Proficiency award for best project work in electrical engineering, IITK, 1993. The Tata Consultancy Services award for best software project at IITK, 1993. Student member IEEE, ISCA. PUBLICATIONS Journal Articles: - Goel V. and Byrne W.J., ``Minimum Bayes-risk automatic speech recognition,'' Computer Speech and Language, Vol. 14(2), pp. 115--135, 2000. - Goel V., Brambrink A.M., Baykal A., and Thakor N.V., ``Dominant frequency analysis reveals brain's response to injury and recovery,'' IEEE Transactions on Biomedical Engineering, Vol. 43(11), pp. 1083--1092, 1996. Selected Conference and Workshop Papers: - Goel V. and Byrne W.J., ``Applications of Minimum Bayes-Risk Decoding to LVCSR,'' LVCSR Workshop, May 16-19, UMCP, College Park, MD, 2000. - Goel V. and Byrne W.J., ``Task dependent loss functions in speech recognition: A-star search over recognition lattices,'' Eurospeech-99, pp. 1243--1246, Budapest, Hungary, 1999. - Goel V. and Byrne W.J., ``Task dependent loss functions in speech recognition: Application to named entity extraction,'' ESCA ETRW Workshop on Accessing Information from Spoken Audio, pp. 49--53, Cambridge, UK, 1999. - Goel V., Byrne W.J., and Khudanpur S., ``LVCSR rescoring with modified loss functions: A decision theoretic perspective,'' ICASSP-98, May 9-12, 1998. - Ganapathiraju A., Goel V., Picone J., et.al., ``Syllable - A promising recognition unit for LVCSR,'' 1997 IEEE Workshop on Automatic Speech Recognition and Understanding, Dec 14-17, 1997. - Corrada A., Doddington G., Ganapathiraju A., Goel V., et.al., ``The syllable team,'' HUB-5 Conversational Speech Recognition Workshop, Nov 4-6, 1997. - Goel V. and Thakor N.V., ``A model of excitotoxic injury to the hippocampal neurons,'' Neural Modeling of Cognitive and Brain Disorders, Jun 8-10, 1995. - Kong X., Goel V., and Thakor N.V., ``Quantification of injury-related EEG signal changes using Itakura distance measure,'' ICASSP-95, May 9-12, 1995. - Goel V., Brambrink A.M., Baykal A., and Thakor N.V., ``Autoregressive modeling of EEG reveals brain's response to injury,'' IEEE/EMBS Annual International Conference, Nov 3-6, 1994. - Goel V., Brambrink A.M., Baykal A., and Thakor N.V., ``Detection of hypoxic-asphyxic insult and recovery using autoregressive modeling of EEG signals,'' American EEG Society Annual Meeting, Sep 19-20, 1994. COMPUTER SKILLS Programming Languages: C, C++, VisualBasic, Pascal. Scripting Languages: Perl, Awk. Toolkits: Matlab, HTK, AT&T FSM Tools. Operatring Systems: Unix, Windows98, WindowsNT. VISA STATUS Indian citizen, at present on F-1 (student visa). REFERENCES Available upon request.