I'm using a GMM model + Time-Delay Neural Network factored (T-DNNf) + RNN language model (which architecture is composed of 2 LSTM layers with 200 cells, 1 tdnn layer and 1 Dense output layer) The goal of this model is to build a better language model and can re-score the result obtained in the first pass decoding.
I have been using this arch during a long time and I wonder if there are new neural networks in Kaldi to use for Acoustic and/or Language model.
Thanks in advance.
I'm using a GMM model + Time-Delay Neural Network factored (T-DNNf) + RNN language model (which architecture is composed of 2 LSTM layers with 200 cells, 1 tdnn layer and 1 Dense output layer) The goal of this model is to build a better language model and can re-score the result obtained in the first pass decoding.
I have been using this arch during a long time and I wonder if there are new neural networks in Kaldi to use for Acoustic and/or Language model. Thanks in advance.