"Hot speech summer" in Brno

Honza Cernocky

SLTC Newsletter, April 2010

Several exciting speech events will take place at Brno University of Technology (BUT), Czech Republic, this summer. All of them are organized or co-organized by BUT Speech@FIT group (headed by Lukas Burget, Honza Cernocky and Hynek Hermansky) at the Faculty of Information Technology of BUT, and are targeted at speaker and speech recognition.

The show will begin with the NIST (U.S. National Institute of Standards and Technology) Speaker Recognition Evaluation (SRE) workshop on the 24th and 25th of June. NIST’s public evaluations provide valuable calibration of the current state of the art, and thus strongly influence the direction of research. Consequently, much research in the speech and language communities is organized around these evaluations. The workshop is the closing event of every evaluation, where the results are announced, and the participants present the details of their systems.

The following event, "Odyssey 2010 – The Speaker and Language Recognition Workshop", held from 28 June – 1 July, is a small but prestigious event in the speaker and language recognition circles. Odyssey is organized as an ISCA Tutorial and Research Workshop and concentrates on the theory and applications of speaker and language recognition for commercial, forensic, and government applications. This workshop in Brno succeeds previous successful events held in Martigny (1994), Avignon (1998), Crete (2001), Toledo (2004), San Juan (2006) and Stellenbosch (2008).

Unlike the previous two events, BOSARIS’10 (Brno Speaker Recognition Summer Workshop 2010) is a 5-week research workshop. Organized around the core of the group that gathered in 2008 at Johns Hopkins summer workshop group “Robust Speaker Recognition over Varying Channels”, the participants will address the issues of inter-session and environmental robustness, and the speed of SRE systems. Although the results of the workshop are likely to be most interesting to the defense and security community, BOSARIS is not a classified research workshop and results will be made available to the public domain. The workshop is lead by Niko Brummer (Agnitio, South Africa), Patrick Kenny (CRIM, Canada) and Lukas Burget (BUT).

Finally, BUT will host several experienced speech researchers for another five weeks for the “KALDI” workshop. The group will be coordinated by Dan Povey (Microsoft), who also headed the 2009 JHU work-group “Low Development Cost, High Quality Speech Recognition for New Languages and Domains”. The KALDI workshop will focus on open and reusable sets of tools to speed up research and development of speech recognition systems. The title of the workshop has something to do with the devotion of some BUT staff to excellent coffee: according to legend, Kaldi was the Ethiopian goat-herder who discovered the coffee plant. Kaldi was also the original name of BUT’s wFST decoder.

If these four events leave you wanting more speech research in the Czech Republic, remember that a year later, ICASSP 2011 will take place in ... Prague!

About the author:

Honza Cernocky is Head of Department of Computer Graphics and Multimedia, Faculty of Information Technology, Brno University of Technology, and managing head of BUT Speech@FIT group. Email: cernocky@fit.vutbr.cz