System and method for speech recognition using deep recurrent neural networks
Deep recurrent neural networks applied to speech recognition. The deep recurrent neural networks (RNNs) are preferably implemented by stacked long short-term memory bidirectional RNNs. The RNNs are trained using end-to-end training with suitable regularisation.
1. A method performed by one or more computers, the method comprising:
receiving an input audio sequence comprising T audio inputs;
processing the input audio sequence using a transcription neural network to generate a sequence of transcription hidden vectors that includes a respective transcription hidden vector for each audio input in the audio sequence; and
generating an output sequence comprising U output labels that represents a transcription of the input audio sequence, wherein each of the output labels is selected from a vocabulary of output labels that comprises a plurality of output labels that each represent a text symbol, and wherein generating an output label at position u in the output sequence comprises:
processing a current output sequence that includes the output labels at positions 1 through u−1 in the output sequence to generate a prediction hidden vector; and
processing the prediction hidden vector and the respective transcription hidden vector for an audio input at a position t in the audio sequence to generate a probability distribution comprising a respective probability for each output label in the vocabulary.
2. The method of claim 1 , wherein generating the output sequence comprises decoding the output sequence using a beam search, and wherein generating the probability distribution is part of the beam search.
3. The method of claim 1 , wherein the transcription neural network is a bi-directional recurrent neural network.
4. The method of claim 3 , wherein the transcription neural network is a bi-directional long short-term memory (LSTM) neural network.
5. The method of claim 4 , wherein each transcription hidden vector is a concatenation of hidden vectors for the corresponding audio input generated by a set of uppermost bidirectional layers in the bi-directional LSTM neural network.
6. The method of claim 1 , wherein the prediction neural network is a uni-directional recurrent neural network.
7. The method of claim 6 , wherein the prediction neural network is a uni-directional long short-term memory (LSTM) neural network.
8. The method of claim 1 , wherein processing the prediction hidden vector and the respective transcription hidden vector for an audio input at a position t in the audio sequence to generate a probability distribution comprising a respective probability for each output label in the vocabulary comprises:
processing the prediction hidden vector and the respective transcription hidden vector using a feedforward neural network to generate the probability distribution.
9. The method of claim 1 , wherein the vocabulary of output labels also includes a blank symbol that represents a non-output.
10. The method of claim 9 , wherein the current output sequence includes the blank symbol followed by the output labels at positions 1 through u−1 in the output sequence.
11. A system comprising one or more computers and one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations comprising:
receiving an input audio sequence comprising T audio inputs;
processing the input audio sequence using a transcription neural network to generate a sequence of transcription hidden vectors that includes a respective transcription hidden vector for each audio input in the audio sequence; and
generating an output sequence comprising U output labels that represents a transcription of the input audio sequence, wherein each of the output labels is selected from a vocabulary of output labels that comprises a plurality of output labels that each represent a text symbol, and wherein generating an output label at position u in the output sequence comprises:
processing a current output sequence that includes the output labels at positions 1 through u−1 in the output sequence to generate a prediction hidden vector; and
processing the prediction hidden vector and the respective transcription hidden vector for an audio input at a position t in the audio sequence to generate a probability distribution comprising a respective probability for each output label in the vocabulary.
12. The system of claim 11 , wherein generating the output sequence comprises decoding the output sequence using a beam search, and wherein generating the probability distribution is part of the beam search.
13. The system of claim 11 , wherein the transcription neural network is a bi-directional recurrent neural network.
14. The system of claim 13 , wherein the transcription neural network is a bi-directional long short-term memory (LSTM) neural network.
15. The system of claim 14 , wherein each transcription hidden vector is a concatenation of hidden vectors for the corresponding audio input generated by a set of uppermost bidirectional layers in the bi-directional LSTM neural network.
16. The system of claim 11 , wherein the prediction neural network is a uni-directional recurrent neural network.
17. The system of claim 16 , wherein the prediction neural network is a uni-directional long short-term memory (LSTM) neural network.
18. The system of claim 11 , wherein processing the prediction hidden vector and the respective transcription hidden vector for an audio input at a position t in the audio sequence to generate a probability distribution comprising a respective probability for each output label in the vocabulary comprises:
processing the prediction hidden vector and the respective transcription hidden vector using a feedforward neural network to generate the probability distribution.
19. The system of claim 11 , wherein the vocabulary of output labels also includes a blank symbol that represents a non-output.
20. One or more non-transitory computer-readable media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
receiving an input audio sequence comprising T audio inputs;
processing the input audio sequence using a transcription neural network to generate a sequence of transcription hidden vectors that includes a respective transcription hidden vector for each audio input in the audio sequence; and
generating an output sequence comprising U output labels that represents a transcription of the input audio sequence, wherein each of the output labels is selected from a vocabulary of output labels that comprises a plurality of output labels that each represent a text symbol, and wherein generating an output label at position u in the output sequence comprises:
processing a current output sequence that includes the output labels at positions 1 through u−1 in the output sequence to generate a prediction hidden vector; and
processing the prediction hidden vector and the respective transcription hidden vector for an audio input at a position t in the audio sequence to generate a probability distribution comprising a respective probability for each output label in the vocabulary.