11th Annual Conference of the International Speech Communication Association

Makuhari, Chiba, Japan
September 26-30. 2010

The AMIDA 2009 Meeting Transcription System

Thomas Hain (1), Lukáš Burget (2), John Dines (3), Philip N. Garner (3), Asmaa El Hannani (1), Marijn Huijbregts (4), Martin Karafiát (2), Mike Lincoln (5), Vincent Wan (1)

(1) University of Sheffield, UK
(2) Brno University of Technology, Czech Republic
(3) Idiap Research Institute, Switzerland
(4) University of Twente, The Netherlands
(5) University of Edinburgh, UK

We present the AMIDA 2009 system for participation in the NIST RT'2009 STT evaluations. Systems for close-talking, far field and speaker attributed STT conditions are described. Improvements to our previous systems are: segmentation and diarisation; stacked bottle-neck posterior feature extraction; fMPE training of acoustic models; adaptation on complete meetings; improvements to WFST decoding; automatic optimisation of decoders and system graphs. Overall these changes gave a 6-13% relative reduction in word error rate while at the same time reducing the real-time factor by a factor of five and using considerably less data for acoustic model training.

Full Paper

Bibliographic reference.  Hain, Thomas / Burget, Lukáš / Dines, John / Garner, Philip N. / Hannani, Asmaa El / Huijbregts, Marijn / Karafiát, Martin / Lincoln, Mike / Wan, Vincent (2010): "The AMIDA 2009 meeting transcription system", In INTERSPEECH-2010, 358-361.