Antti-Veikko Rosti, PhD (Cantab)

Profile

I am a Scientist in the Speech & Language Processing Department at Raytheon BBN Technologies. My research interests are in signal processing and machine learning with applications to audio, speech and language processing. In my PhD research, I investigated a set of statistical models applied to acoustic modeling for speech recognition. My current work is focused on speech recognition and machine translation in the Autonomous Global Integrated Language Exploitation (AGILE) project (part of the DARPA GALE program). I am the main contributor to BBN machine translation system combination in the GALE, WMT08, WMT09, and NIST MT09 evaluations. Nowadays, all my spare time is spent raising my little pumpkins, Aliisa and Aleksi.

Curriculum Vitae

Education

• PhD in Engineering, Wolfson College, University of Cambridge, 2004
  • Research on acoustic modeling using generalized linear Gaussian models in the Speech Research Group, Machine Intelligence Laboratory
  • Thesis supervised by Dr Mark Gales
• MSc in Engineering "with distinction", Department of Information Technology, Tampere University of Technology, 1999
  • Main subject Signal Processing, Subsidiary subject Software Systems
  • Thesis supervised by Prof Visa Koivunen
• Matriculation Examination, The Sixth Form College of Jyväskylä Lyseo, 1994

Experience

• Postdoctoral Researcher in the Human Language Technologies Group at IBM T.J. Watson Research Center
  • Active learning for natural language call routing
  • Phrase-based machine translation for speech-to-speech translation
  • Assist in data collection efforts for the DARPA TransTac program
• Research Associate in the HTK Rich Audio Transcription project (part of the DARPA EARS program)
  • Automatic segmentation for conversational telephone speech
  • HTK software development
• Research Scientist in the Audio Research Group, Institute of Signal Processing, Tampere University of Technology
  • SpeechDat-Car
  • SpeechDat(II)
  • Classification of acoustic and communication signals

Selected Publications

• List of Publications
• Machine Translation
  • S. Matsoukas, A.-V.I. Rosti, and B. Zhang . Discriminative Corpus Weight Estimation for Machine Translation. In Proc. EMNLP, pp. 708-717, 2009.
  • A.-V.I. Rosti, B. Zhang, S. Matsoukas, and R. Schwartz. Incremental Hypothesis Alignment with Flexible Matching for Building Confusion Networks: BBN System Description for WMT09 System Combination Task. In Proc. EACL/WMT, pp. 61-65, 2009.
  • A.-V.I. Rosti, B. Zhang, S. Matsoukas, and R. Schwartz. Incremental Hypothesis Alignment for Building Confusion Networks with Application to Machine Translation System Combination. In Proc. ACL/WMT, pp. 183-186, 2008.
  • A.-V.I. Rosti, S. Matsoukas, and R. Schwartz. Improved word-level system combination for machine translation. In Proc. ACL, pp. 312-319, 2007.
  • A.-V.I. Rosti, N.F. Ayan, B. Xiang, S. Matsoukas, R. Schwartz, and B.J. Dorr. Combining outputs from multiple machine translation systems. In Proc. NAACL-HLT, pp. 228-235, 2007. Best paper award.
  • Y. Gao, B. Zhou, L. Gu, R. Sarikaya, H.-K. Kuo, A.-V.I. Rosti, M.A. Afify, and W. Zhu. IBM Mastor: Multilingual automatic speech-to-speech translator. In Proc. ICASSP, vol. 5, pp. 1205-1208, 2006.
• Speech Recognition
  • A.-V.I. Rosti. Linear Gaussian Models for Speech Recognition. PhD thesis, University of Cambridge, 2004.
  • A.-V.I. Rosti and M.J.F. Gales. Rao-Blackwellised Gibbs sampling for switching linear dynamical systems. In Proc. ICASSP, vol. 1, pp. 809-812, 2004.
  • A.-V.I. Rosti and M.J.F. Gales. Factor analysed hidden Markov models for speech recognition. Computer Speech and Language, 18(2):181-200, 2004.
  • A.-V.I. Rosti and M.J.F. Gales. Switching linear dynamical systems for speech recognition. Technical Report CUED/F-INFENG/TR.461, Cambridge University Engineering Department, 2003.
  • A.-V.I. Rosti and M.J.F. Gales. Factor analysed hidden Markov models. In Proc. ICASSP, vol. 1, pp. 949-952, 2002.
  • A.-V.I. Rosti and M.J.F. Gales. Generalised linear Gaussian models. Technical Report CUED/F-INFENG/TR.420, Cambridge University Engineering Department, 2001.
• Statistical Signal Processing and Classification
  • A.-V. Rosti and V. Koivunen. Classification of MFSK modulated signals using the mean of complex envelope. In Proc. EUSIPCO, vol. 1, pp. 581-584, 2000.
  • A.-V. Rosti. Statistical methods in modulation classification. Master's thesis, Tampere University of Technology, 1999.
• Others
  • Introductory notes (1999): 1-bit A/D and D/A Converters
  • Signal Processing Project (1998): Finnish Spoken Digit Recognizer