The result is a combination of the application of state-of-the-art speech recognition methods such as simple dialect adaptation for a Time-Delay Neural Network (TDNN) acoustic model (-27% errors compared to the baseline), Recurrent Neural Network Language Model (RNNLM) rescoring (an additional -5%), and system combination with Minimum Bayes Risk (MBR) decoding (yet another -10%). We also explored the use of morph and character language models, which was particularly beneficial in providing a rich pool of systems for the MBR decoding.
|Title of host publication||Automatic Speech Recognition and Understanding (ASRU), IEEE Workshop on|
|Publication status||Published - 2018|
|MoE publication type||A4 Article in a conference publication|
|Event||IEEE Automatic Speech Recognition and Understanding Workshop - Okinawa, Japan|
Duration: 16 Dec 2017 → 20 Dec 2017
|Workshop||IEEE Automatic Speech Recognition and Understanding Workshop|
|Period||16/12/2017 → 20/12/2017|
FingerprintDive into the research topics of 'Aalto system for the 2017 Arabic multi-genre broadcast challenge'. Together they form a unique fingerprint.
Smit, P. (Recipient), Gangireddy, S. (Recipient), Enarvi, S. (Recipient), Virpioja, S. (Recipient) & Kurimo, Mikko (Recipient), 2017
Prize: Invitation or ranking in competition