LSTM-CRF-和-CTC算法详解-(纯英版).pptxVIP

Speech recognition and NER problems with recurrent neural networks;RNN;RNN and CNN;LSTM;Comparing the fully connected layers, the parameters in each LSTM neuron will be more than at lease 4 times more. So there is GRU, and you can consider using GRU when LSTM is facing a risk of overfitting In other articles, sigmoid activation is sometimes used instead of tanh activation. It overcomes the gradient vanish problem, but still cant solve the problem of gradient explosion;Further;BRNNs;Deep bidirectional RNNs;Structure learning HMMs;X: Y: start: John saw the tree. end;From HMMs to CRFs;LSTMs-CRFs;P(x, y) for CRFs;?;CRF-gradient ascent;?;?;Connectionist Temporal Classification (CTC) ;?;The CTC network predicts only the sequence of phonemes (typically as a series of spikes, separated by ‘blanks’). The choice of labelling can be read directly from the CTC outputs (follow the spikes).And the predictions of the framewise network must be post-processed before use.;?;Prefix search decoding;The CTC Forward-Backward Algorithm;?;result;;Thank you