You can improve the speech-to-text success rate by:
Adapting the acoustic model to suit the specific audio type, quality, dialect, or collection of speakers.
From the 10.8 release of HPE IDOL Speech Server, many language packs uses Deep Neural Network acoustic models (DNN files). These models cannot be adapted but typically outperform even adapted GMM acoustic models. For this reason, does not generally recommend acoustic adaptation any more. For more information, see Adapt Acoustic Models.
|