Google Inc.

Zhiyun Lu

Yu Zhang

Liangliang Cao

Chao Zhang

Thibault Doutre

Tara N Sainath

Bo Li

Ruoming Pang

Arun Narayanan

Chung-Cheng Chiu

Wei Han

Ananya Misra

Shuo-Yiin Chang

Yongqiang Wang

Min Ma

Yanwei Pan

A method includes obtaining a corpus of unlabeled training data including a plurality of spoken utterances, each corresponding spoken utterance of the plurality of spoken utterances includes audio data characterizing the corresponding spoken utterance. The method also includes receiving a target domain. The method also includes selecting, using a contrastive data selection model, a subset of the utterances from the corpus of unlabeled training data that correspond to the target domain. The method includes training an automatic speech recognition (ASR) model on the subset of utterances.

Unsupervised data selection via discrete speech representation for automatic speech recognition

A method includes obtaining a set of training samples, wherein each training sample includes a corresponding sequence of speech segments corresponding to a training utterance and a corresponding sequence of ground-truth transcriptions for the sequence of speech segments, and wherein each ground-truth transcription includes a start time and an end time of a corresponding speech segment. For each training sample in the set of training samples, the method includes processing, using a speech recognition model, the corresponding sequence of speech segments to obtain one or more speech recognition hypotheses for the training utterance; and, for each speech recognition hypothesis obtained for the training utterance, identifying a respective number of word errors relative to the corresponding sequence of ground-truth transcriptions. The method trains the speech recognition model to minimize word error rate based on the respective number of word errors identified for each speech recognition hypothesis obtained for the training utterance.

Training for long-form speech recognition

A speech recognition model includes an encoder network, a prediction network, and a joint network. The encoder network is configured to receive a sequence of acoustic frames characterizing an input utterance; and generate, at each of a plurality of output steps, a higher order feature representation for a corresponding acoustic frame in the sequence of acoustic frames. The prediction network is configured to: receive a sequence of non-blank symbols output by a final Softmax layer; and generate, at each of the plurality of output steps, a dense representation. The joint network is configured to generate, at each of the plurality of output steps based on the higher order feature representation and the dense representation, a probability distribution over possible speech recognition hypotheses. The joint network includes a stack of gating and bilinear pooling to fuse the dense representation and the higher order feature representation.

Fusion of acoustic and text representations in RNN-T

A method for training a streaming automatic speech recognition student model includes receiving a plurality of unlabeled student training utterances. The method also includes, for each unlabeled student training utterance, generating a transcription corresponding to the respective unlabeled student training utterance using a plurality of non-streaming automated speech recognition (ASR) teacher models. The method further includes distilling a streaming ASR student model from the plurality of non-streaming ASR teacher models by training the streaming ASR student model using the plurality of unlabeled student training utterances paired with the corresponding transcriptions generated by the plurality of non-streaming ASR teacher models.

Streaming automatic speech recognition with non-streaming model distillation

Growing community of inventors

Brooklyn, NY, United States of America

Zhiyun Lu