Publication Details
The AMIDA 2009 Meeting Transcription System
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Dines John
Garner Phillip
El Hannani Asmaa
Huijbregts Marijn
Karafiát Martin, Ing., Ph.D. (DCGM)
Lincoln Mike
Wan Vincent
speech recognition, meeting transcription
The paper is on systems for close-taking, far field and speaker attributed STT conditions. The system was used at participation in the NIST RT'2009 STT evaluations.
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.
@inproceedings{BUT35027,
author="Thomas {Hain} and Lukáš {Burget} and John {Dines} and Phillip {Garner} and Asmaa {El Hannani} and Marijn {Huijbregts} and Martin {Karafiát} and Mike {Lincoln} and Vincent {Wan}",
title="The AMIDA 2009 Meeting Transcription System",
booktitle="Proceedings of the 11th Annual Conference of the International Speech Communication Association (INTERSPEECH 2010)",
year="2010",
journal="Proceedings of Interspeech",
volume="2010",
number="9",
pages="358--361",
publisher="International Speech Communication Association",
address="Makuhari, Chiba",
isbn="978-1-61782-123-3",
issn="1990-9772",
url="http://www.fit.vutbr.cz/research/groups/speech/publi/2010/burget_interspeech2010_IS100781.pdf"
}