Abstract: Audio signals of speech may be processed using an acoustic model. An acoustic model may be implemented with multiple streams of processing where different streams perform processing using different dilation rates. For example, a first stream may process features of the audio signal with one or more convolutional neural network layers having a first dilation rate, and a second stream may process features of the audio signal with one or more convolutional neural network layers having a second dilation rate. Each stream may compute a stream vector, and the stream vectors may be combined to a vector of speech unit scores, where the vector of speech unit scores provides information about the acoustic content of the audio signal. The vector of speech unit scores may be used for any appropriate application of speech, such as automatic speech recognition.

Abstract: A sequence processing method can be applied to a graphics processor (GPU), and include: determining a sequence to be processed, which has an irregular tensor data structure; determining data structure information in the sequence to be processed, where the data structure information includes tensor dimensions and element information in tensors of each dimension; converting the irregular tensor data structure into a regular tensor data structure based on the tensor dimension and the element information; and processing the sequence to be processed based on the regular tensor data structure. The sequence with irregular tensor data structure can be processed on the GPU, so as to optimize the ability of GPU to process the sequence, speed up the processing process, and improve the efficiency of GPU to process the sequence.

Abstract: Audio signals of speech may be processed using an acoustic model. An acoustic model may be implemented with multiple streams of processing where different streams perform processing using different dilation rates. For example, a first stream may process features of the audio signal with one or more convolutional neural network layers having a first dilation rate, and a second stream may process features of the audio signal with one or more convolutional neural network layers having a second dilation rate. Each stream may compute a stream vector, and the stream vectors may be combined to a vector of speech unit scores, where the vector of speech unit scores provides information about the acoustic content of the audio signal. The vector of speech unit scores may be used for any appropriate application of speech, such as automatic speech recognition.

Abstract: Subspace speech adaptation may be utilized for facilitating the recognition of speech containing short utterances. Speech training data may be received in a speech model by a computer. A first matrix may be determined for preconditioning speech statistics based on the speech training data. A second matrix may be determined for representing a basis for the speech to be recognized. A set of basis matrices may then be determined from the first matrix and the second matrix. Speech test data including a short utterance may then be received by the computer. The computer may then apply the set of basis matrices to the speech test data to produce a transcription. The transcription may represent speech recognition of the short utterance.

Abstract: Subspace speech adaptation may be utilized for facilitating the recognition of speech containing short utterances. Speech training data may be received in a speech model by a computer. A first matrix may be determined for preconditioning speech statistics based on the speech training data. A second matrix may be determined for representing a basis for the speech to be recognized. A set of basis matrices may then be determined from the first matrix and the second matrix. Speech test data including a short utterance may then be received by the computer. The computer may then apply the set of basis matrices to the speech test data to produce a transcription. The transcription may represent speech recognition of the short utterance.

Abstract: A method of optimizing a function of a parameter includes associating, with an objective function for initial value of parameters, an auxiliary function of parameters that could be optimized computationally more efficiently than an original objective function, obtaining parameters that are optimum for the auxiliary function, obtaining updated parameters by taking a weighted sum of the optimum of the auxiliary function and initial model parameters.

Abstract: A method of optimizing a function of a parameter includes associating, with an objective function for initial value of parameters, an auxiliary function of parameters that could be optimized computationally more efficiently than an original objective function, obtaining parameters that are optimum for the auxiliary function, obtaining updated parameters by taking a weighted sum of the optimum of the auxiliary function and initial model parameters.