Vlasios Doumpiotis Ph.D Center for Language and Speech Processing, Johns Hopkins University, 320 Barton Hall, 3400 N. Charles Street, Baltimore, MD 21218 Email: vlasios@jhu.edu Home Phone: (781)-929-2062 Current Address: Vlasios Doumpiotis 54 Tillotson Road Suite #2 Needham MA 02494 RESEARCH INTERESTS: Statistical Methods for Speech Recognition and Natural Language. Acoustic and Language Modeling. Adaptation and Pronunciation Modeling. Discriminative Training/Adaptation. Statistical Modeling and Learning. EDUCATION: Ph.D., (Jan 2000 - Sep 2004), Electrical and Computer Engineering, Johns Hopkins University Advisor: Dr. William J. Byrne. M.S.E., (Sep 1998 - Jan 2000), Electrical and Computer Engineering, Johns Hopkins University. Advisor: Dr. Frederick Jelinek. GPA: 3.80/4.0 B.E., (August 93 - Sep 98), Science degree in Electronic and Computer Engineering from the Technical University of Crete. Advisor: Dr. Vassilios Digalakis. RESEARCH: Research Assistant, Jan 2000 - Sep 2004 at the Center for Language and Speech Processing(CLSP), Johns Hopkins University. * Developed Discriminative Linear Transforms for Speaker adaptation in large vocabulary continuous speech recognition tasks. * Studied unsupervised speaker adaptation techniques based on Discriminative Training and Eigen Transforms. * Studied Lattice Segmentation procedures used for Minimun Bayes Risk discriminative training in order to estimate the HMM parameters so as to reduce the overall risk over the training data * Contributed significantly to the acoustic modeling training of the Johns Hopkins University large vocabulary conversational speech transcription systems for the NIST 2002 Rich Transcription evaluation. TEACHING: Dept of Electrical and Computer Engineering, Johns Hopkins University * Teaching Assistant for Electric Circuits (Sept 1998 - Dec1999) (Sept 2000 -Dec 2000) * Teaching Assistant for Digital Signal Fundamentals (Jan 1999 - May1999) (Jan 2000 -May 2000) PUBLICATIONS Thesis title: "Discriminative Training for Speaker Adaptation and Minimum Bayes Risk Estimation in Large Vocabulary Speech Recognition." Conference Papers * V. Digalakis, V. Doumpiotis, S. Tsakalidis. "On the integration of dialect and speaker adaptation in a multi-dialect speech recognition system," EUSIPCO, Rhodes, Greece, September 1998. * S. Tsakalidis, V. Doumpiotis and W. Byrne. "Dicriminative Linear Transforms for Feature Normalization and Speaker Adaptation in HMM Estimation," ICSLP, Denver, 2002. * V. Doumpiotis, S. Tsakalidis and W. Byrne. "Discriminative Training for Segmental Minimum Bayes Risk Decoding," ICASSP,Hong Kong, 2003 * V. Doumpiotis, S. Tsakalidis and W. Byrne. "Lattice Segmentation and Minimum Bayes Risk Discriminative Training," EUROSPEECH,Geneva, 2003 * V. Doumpiotis, Y. Deng. "EIGENSPACE-BASED MLLR WITH SPEAKER ADAPTIVE TRAINING IN LARGE VOCABULARY CONVERSATIONAL SPEECH RECOGNITION" ICASSP, Montreal, 2004 * V. Doumpiotis and W. Byrne. "Pinched Lattice Minimum Bayes Risk Discriminative Training for Large Vocabulary Continuous Speech Recognition" ICSLP04 South Korea. Journal Papers * V. Doumpiotis and W. Byrne. Lattice segmentation and minimum Bayes risk discriminative training for large vocabulary continuous speech recognition. Speech Communication. Accepted, in revision. * S. Tsakalidis, V. Doumpiotis and W. Byrne. "Discriminative Linear Transforms for Feature Normalization and Speaker Adaptation in HMM Estimation," IEEE Transactions on Speech and Audio Processing, 13(3), May 2005. NIST Evaluation Workshops and Technical Reports * W.Byrne, V.Doumpiotis, S.Kumar, S.Tsakalidis, and V.Venkataramani, ``The JHU 2002 large vocabulary speech recognition system,'' in Proceedings of the NIST RT-02 Workshop. NIST, 2002, Available at HREF="http://www.clsp.jhu.edu/research/rteval/". Research Proposal: The Johns Hopkins University ECE Department August 2003 * "Discriminative Training for Speaker Adaptation and Minimum Risk Estimation in Large Vocabulary Conversational Speech Recognition" Technical Reports * S. Tsakalidis, V. Doumpiotis and W. Byrne. "Discriminative Linear Transforms for Feature Normalization and Speaker Adaptation in HMM Estimation," CLSP Technical Report, No.47. Honors and Awards: o December 1999-2001 Received award from the Greek Ministry of Education for outstanding academic performance among graduate students. o January 1997 Received award among the 5 Best Students in Technical University of Crete Electrical Computer Engineering Department. o March 1999 Award of Excellence in Telecommunications 1999 for my thesis by Ericsson Hellas (speech recognition using normalization techniques VTLN). Work Experience: Current Employer: Since September 2004, I have joined the ASR group at Escription Inc(Needham MA). Working on discriminative acoustic model training and improving the Sphinx decoder. Summer Intern: L&H, Boston, MA. (Now Scansoft), May 1999 - Aug 1999. Project: Speaker Normalization and Vocal Tract Length Normalization. Supervisor: George Zavaliagkos. Incoporated VTN-based speaker normalization schemes in L&H(Scansoft) recognizer. Obtained substantial WER improvements in several dialog modules of the recognizer. L&H, Boston, MA. (Now Scansoft), May 2000 - Aug 2000 Project: Discriminative Linear tranforms for speaker normalization. Supervisor: George Zavaliagkos. RELEVANT COURSEWORK: Speech and Language: Introduction to Large Vocabulary Continuous Speech Recognition Information Extraction from Text and Speech Natural Language Processing Information Retrieval and Web Agents Machine Learning and Statistical Theory: Information Theory Statistical Theory COMPUTER SKILLS: Programming Languages: C, C++, Perl, Matlab. Toolkits: HTK, AT&T FSM Tools, Sphinx Decoder. Operating Systems: Unix, Windows, Mac OS X. VISA STATUS: Greek citizen, H-1 (visa).