could i used the DFSMN language model to provide the service

alumae / kaldi-gstreamer-server

Real-time full-duplex speech recognition server, based on the Kaldi toolkit and the GStreamer framwork.

BSD 2-Clause "Simplified" License

1.06k stars 341 forks source link

could i used the DFSMN language model to provide the service #146

Closed AlexPeng19 closed 5 years ago

AlexPeng19 commented 5 years ago

Dear Tanel : You know the alibaba/Alibaba-MIT-Speech project on github used the DFSMN language model based on KALDI project, which improved the Speech recognition recognition rate into 96%. so i have trained the language model, could i know is it compatible for our project to provide service based on that model. if it is, what changes should i made. @alumae Looking forward your response.

alumae commented 5 years ago

I am afraid kaldi-gstreamer-server is not compatible with the Alibaba models as it requires heavy patching of Kaldi.