Jasha Droppo

Xin Zhang, Iván Vallés-Pérez, Andreas Stolcke et al. · amazon-science

Stuttering is a speech disorder where the natural flow of speech is interrupted by blocks, repetitions or prolongations of syllables, words and phrases. The majority of existing automatic speech recognition (ASR) interfaces perform poorly on utterances with stutter, mainly due to lack of matched training data. Synthesis of speech with stutter thus presents an opportunity to improve ASR for this type of speech. We describe Stutter-TTS, an end-to-end neural text-to-speech model capable of synthesizing diverse types of stuttering utterances. We develop a simple, yet effective prosody-control strategy whereby additional tokens are introduced into source text during training to represent specific stuttering characteristics. By choosing the position of the stutter tokens, Stutter-TTS allows word-level control of where stuttering occurs in the synthesized utterance. We are able to synthesize stutter events with high accuracy (F1-scores between 0.63 and 0.84, depending on stutter type). By fine-tuning an ASR model on synthetic stuttered speech we are able to reduce word error by 5.7% relative on stuttered utterances, with only minor (<0.2% relative) degradation for fluent utterances.

Gopinath Chennupati, Milind Rao, Gurpreet Chadha et al.

Incremental learning is one paradigm to enable model building and updating at scale with streaming data. For end-to-end automatic speech recognition (ASR) tasks, the absence of human annotated labels along with the need for privacy preserving policies for model building makes it a daunting challenge. Motivated by these challenges, in this paper we use a cloud based framework for production systems to demonstrate insights from privacy preserving incremental learning for automatic speech recognition (ILASR). By privacy preserving, we mean, usage of ephemeral data which are not human annotated. This system is a step forward for production levelASR models for incremental/continual learning that offers near real-time test-bed for experimentation in the cloud for end-to-end ASR, while adhering to privacy-preserving policies. We show that the proposed system can improve the production models significantly(3%) over a new time period of six months even in the absence of human annotated labels with varying levels of weak supervision and large batch sizes in incremental learning. This improvement is 20% over test sets with new words and phrases in the new time period. We demonstrate the effectiveness of model building in a privacy-preserving incremental fashion for ASR while further exploring the utility of having an effective teacher model and use of large batch sizes.

Guruprasad V Ramesh, Gopinath Chennupati, Milind Rao et al.

Federated Learning (FL) is a privacy-preserving paradigm, allowing edge devices to learn collaboratively without sharing data. Edge devices like Alexa and Siri are prospective sources of unlabeled audio data that can be tapped to learn robust audio representations. In this work, we bring Self-supervised Learning (SSL) and FL together to learn representations for Automatic Speech Recognition respecting data privacy constraints. We use the speaker and chapter information in the unlabeled speech dataset, Libri-Light, to simulate non-IID speaker-siloed data distributions and pre-train an LSTM encoder with the Contrastive Predictive Coding framework with FedSGD. We show that the pre-trained ASR encoder in FL performs as well as a centrally pre-trained model and produces an improvement of 12-15% (WER) compared to no pre-training. We further adapt the federated pre-trained models to a new language, French, and show a 20% (WER) improvement over no pre-training.

Aparna Khare, Minhua Wu, Saurabhchand Bhati et al.

Contrastive Predictive Coding (CPC) is a representation learning method that maximizes the mutual information between intermediate latent representations and the output of a given model. It can be used to effectively initialize the encoder of an Automatic Speech Recognition (ASR) model. We present a novel modification of CPC called Guided Contrastive Predictive Coding (GCPC). Our proposed method maximizes the mutual information between representations from a prior-knowledge model and the output of the model being pre-trained, allowing prior knowledge injection during pre-training. We validate our method on 3 ASR tasks: German, French and English. Our method outperforms CPC pre-training on all three datasets, reducing the Word Error Rate (WER) by 4.44%, 6.55% and 15.43% relative on the German, French and English (Librispeech) tasks respectively, compared to training from scratch, while CPC pre-training only brings 2.96%, 1.01% and 14.39% relative WER reduction respectively.

Hua Shen, Yuguang Yang, Guoli Sun et al.

Modern speaker verification models use deep neural networks to encode utterance audio into discriminative embedding vectors. During the training process, these networks are typically optimized to differentiate arbitrary speakers. This learning process biases the learning of fine voice characteristics towards dominant demographic groups, which can lead to an unfair performance disparity across different groups. This is observed especially with underrepresented demographic groups sharing similar voice characteristics. In this work, we investigate the fairness of speaker verification models on controlled datasets with imbalanced gender distributions, providing direct evidence that model performance suffers for underrepresented groups. To mitigate this disparity we propose the group-adapted fusion network (GFN) architecture, a modular architecture based on group embedding adaptation and score fusion. We show that our method alleviates model unfairness by improving speaker verification both overall and for individual groups. Given imbalanced group representation in training, our proposed method achieves overall equal error rate (EER) reduction of 9.6% to 29.0% relative, reduces minority group EER by 13.7% to 18.6%, and results in 20.0% to 25.4% less EER disparity, compared to baselines. The approach is applicable to other types of training data skew in speaker recognition systems.

I-Fan Chen, Brian King, Jasha Droppo

In this paper, we propose an approach to quantitatively analyze impacts of different training label errors to RNN-T based ASR models. The result shows deletion errors are more harmful than substitution and insertion label errors in RNN-T training data. We also examined label error impact mitigation approaches on RNN-T and found that, though all the methods mitigate the label-error-caused degradation to some extent, they could not remove the performance gap between the models trained with and without the presence of label errors. Based on the analysis results, we suggest to design data pipelines for RNN-T with higher priority on reducing deletion label errors. We also find that ensuring high-quality training labels remains important, despite of the existence of the label error mitigation approaches.

Amin Fazel, Wei Yang, Yulan Liu et al.

End-to-end (E2E) automatic speech recognition (ASR) models have recently demonstrated superior performance over the traditional hybrid ASR models. Training an E2E ASR model requires a large amount of data which is not only expensive but may also raise dependency on production data. At the same time, synthetic speech generated by the state-of-the-art text-to-speech (TTS) engines has advanced to near-human naturalness. In this work, we propose to utilize synthetic speech for ASR training (SynthASR) in applications where data is sparse or hard to get for ASR model training. In addition, we apply continual learning with a novel multi-stage training strategy to address catastrophic forgetting, achieved by a mix of weighted multi-style training, data augmentation, encoder freezing, and parameter regularization. In our experiments conducted on in-house datasets for a new application of recognizing medication names, training ASR RNN-T models with synthetic audio via the proposed multi-stage training improved the recognition performance on new application by more than 65% relative, without degradation on existing general applications. Our observations show that SynthASR holds great promise in training the state-of-the-art large-scale E2E ASR models for new applications while reducing the costs and dependency on production data.

Rupak Vignesh Swaminathan, Brian King, Grant P. Strimel et al.

We propose a simple yet effective method to compress an RNN-Transducer (RNN-T) through the well-known knowledge distillation paradigm. We show that the transducer's encoder outputs naturally have a high entropy and contain rich information about acoustically similar word-piece confusions. This rich information is suppressed when combined with the lower entropy decoder outputs to produce the joint network logits. Consequently, we introduce an auxiliary loss to distill the encoder logits from a teacher transducer's encoder, and explore training strategies where this encoder distillation works effectively. We find that tandem training of teacher and student encoders with an inplace encoder distillation outperforms the use of a pre-trained and static teacher transducer. We also report an interesting phenomenon we refer to as implicit distillation, that occurs when the teacher and student encoders share the same decoder. Our experiments show 5.37-8.4% relative word error rate reductions (WERR) on in-house test sets, and 5.05-6.18% relative WERRs on LibriSpeech test sets.

Jasha Droppo, Oguz Elibol

There is a recent trend in machine learning to increase model quality by growing models to sizes previously thought to be unreasonable. Recent work has shown that autoregressive generative models with cross-entropy objective functions exhibit smooth power-law relationships, or scaling laws, that predict model quality from model size, training set size, and the available compute budget. These scaling laws allow one to choose nearly optimal hyper-parameters given constraints on available training data, model parameter count, or training computation budget. In this paper, we demonstrate that acoustic models trained with an auto-predictive coding loss behave as if they are subject to similar scaling laws. We extend previous work to jointly predict loss due to model size, to training set size, and to the inherent "irreducible loss" of the task. We find that the scaling laws accurately match model performance over two orders of magnitude in both model size and training set size, and make predictions about the limits of model performance.

Iván Vallés-Pérez, Julian Roth, Grzegorz Beringer et al.

Text-to-speech systems recently achieved almost indistinguishable quality from human speech. However, the prosody of those systems is generally flatter than natural speech, producing samples with low expressiveness. Disentanglement of speaker id and prosody is crucial in text-to-speech systems to improve on naturalness and produce more variable syntheses. This paper proposes a new neural text-to-speech model that approaches the disentanglement problem by conditioning a Tacotron2-like architecture on flow-normalized speaker embeddings, and by substituting the reference encoder with a new learned latent distribution responsible for modeling the intra-sentence variability due to the prosody. By removing the reference encoder dependency, the speaker-leakage problem typically happening in this kind of systems disappears, producing more distinctive syntheses at inference time. The new model achieves significantly higher prosody variance than the baseline in a set of quantitative prosody features, as well as higher speaker distinctiveness, without decreasing the speaker intelligibility. Finally, we observe that the normalized speaker embeddings enable much richer speaker interpolations, substantially improving the distinctiveness of the new interpolated speakers.

Gokce Keskin, Minhua Wu, Brian King et al.

Automatic speech recognition (ASR) models are typically designed to operate on a single input data type, e.g. a single or multi-channel audio streamed from a device. This design decision assumes the primary input data source does not change and if an additional (auxiliary) data source is occasionally available, it cannot be used. An ASR model that operates on both primary and auxiliary data can achieve better accuracy compared to a primary-only solution; and a model that can serve both primary-only (PO) and primary-plus-auxiliary (PPA) modes is highly desirable. In this work, we propose a unified ASR model that can serve both modes. We demonstrate its efficacy in a realistic scenario where a set of devices typically stream a single primary audio channel, and two additional auxiliary channels only when upload bandwidth allows it. The architecture enables a unique methodology that uses both types of input audio during training time. Our proposed approach achieves up to 12.5% relative word-error-rate reduction (WERR) compared to a PO baseline, and up to 16.0% relative WERR in low-SNR conditions. The unique training methodology achieves up to 2.5% relative WERR compared to a PPA baseline.

Swayambhu Nath Ray, Minhua Wu, Anirudh Raju et al.

Comprehending the overall intent of an utterance helps a listener recognize the individual words spoken. Inspired by this fact, we perform a novel study of the impact of explicitly incorporating intent representations as additional information to improve a recurrent neural network-transducer (RNN-T) based automatic speech recognition (ASR) system. An audio-to-intent (A2I) model encodes the intent of the utterance in the form of embeddings or posteriors, and these are used as auxiliary inputs for RNN-T training and inference. Experimenting with a 50k-hour far-field English speech corpus, this study shows that when running the system in non-streaming mode, where intent representation is extracted from the entire utterance and then used to bias streaming RNN-T search from the start, it provides a 5.56% relative word error rate reduction (WERR). On the other hand, a streaming system using per-frame intent posteriors as extra inputs for the RNN-T ASR system yields a 3.33% relative WERR. A further detailed analysis of the streaming system indicates that our proposed method brings especially good gain on media-playing related intents (e.g. 9.12% relative WERR on PlayMusicIntent).

Bhargav Pulugundla, Yang Gao, Brian King et al.

Attention-based beamformers have recently been shown to be effective for multi-channel speech recognition. However, they are less capable at capturing local information. In this work, we propose a 2D Conv-Attention module which combines convolution neural networks with attention for beamforming. We apply self- and cross-attention to explicitly model the correlations within and between the input channels. The end-to-end 2D Conv-Attention model is compared with a multi-head self-attention and superdirective-based neural beamformers. We train and evaluate on an in-house multi-channel dataset. The results show a relative improvement of 3.8% in WER by the proposed model over the baseline neural beamformer.

Samik Sadhu, Di He, Che-Wei Huang et al.

Wav2vec-C introduces a novel representation learning technique combining elements from wav2vec 2.0 and VQ-VAE. Our model learns to reproduce quantized representations from partially masked speech encoding using a contrastive loss in a way similar to Wav2vec 2.0. However, the quantization process is regularized by an additional consistency network that learns to reconstruct the input features to the wav2vec 2.0 network from the quantized representations in a way similar to a VQ-VAE model. The proposed self-supervised model is trained on 10k hours of unlabeled data and subsequently used as the speech encoder in a RNN-T ASR model and fine-tuned with 1k hours of labeled data. This work is one of only a few studies of self-supervised learning on speech tasks with a large volume of real far-field labeled data. The Wav2vec-C encoded representations achieves, on average, twice the error reduction over baseline and a higher codebook utilization in comparison to wav2vec 2.0

Milind Rao, Pranav Dheram, Gautam Tiwari et al.

Spoken language understanding (SLU) systems extract transcriptions, as well as semantics of intent or named entities from speech, and are essential components of voice activated systems. SLU models, which either directly extract semantics from audio or are composed of pipelined automatic speech recognition (ASR) and natural language understanding (NLU) models, are typically trained via differentiable cross-entropy losses, even when the relevant performance metrics of interest are word or semantic error rates. In this work, we propose non-differentiable sequence losses based on SLU metrics as a proxy for semantic error and use the REINFORCE trick to train ASR and SLU models with this loss. We show that custom sequence loss training is the state-of-the-art on open SLU datasets and leads to 6% relative improvement in both ASR and NLU performance metrics on large proprietary datasets. We also demonstrate how the semantic sequence loss training paradigm can be used to update ASR and SLU models without transcripts, using semantic feedback alone.

Daniel Korzekwa, Roberto Barra-Chicote, Szymon Zaporowski et al.

This paper describes two novel complementary techniques that improve the detection of lexical stress errors in non-native (L2) English speech: attention-based feature extraction and data augmentation based on Neural Text-To-Speech (TTS). In a classical approach, audio features are usually extracted from fixed regions of speech such as the syllable nucleus. We propose an attention-based deep learning model that automatically derives optimal syllable-level representation from frame-level and phoneme-level audio features. Training this model is challenging because of the limited amount of incorrect stress patterns. To solve this problem, we propose to augment the training set with incorrectly stressed words generated with Neural TTS. Combining both techniques achieves 94.8% precision and 49.2% recall for the detection of incorrectly stressed words in L2 English speech of Slavic and Baltic speakers.

Jinxi Guo, Gautam Tiwari, Jasha Droppo et al.

In this work, we propose a novel and efficient minimum word error rate (MWER) training method for RNN-Transducer (RNN-T). Unlike previous work on this topic, which performs on-the-fly limited-size beam-search decoding and generates alignment scores for expected edit-distance computation, in our proposed method, we re-calculate and sum scores of all the possible alignments for each hypothesis in N-best lists. The hypothesis probability scores and back-propagated gradients are calculated efficiently using the forward-backward algorithm. Moreover, the proposed method allows us to decouple the decoding and training processes, and thus we can perform offline parallel-decoding and MWER training for each subset iteratively. Experimental results show that this proposed semi-on-the-fly method can speed up the on-the-fly method by 6 times and result in a similar WER improvement (3.6%) over a baseline RNN-T model. The proposed MWER training can also effectively reduce high-deletion errors (9.2% WER-reduction) introduced by RNN-T models when EOS is added for endpointer. Further improvement can be achieved if we use a proposed RNN-T rescoring method to re-rank hypotheses and use external RNN-LM to perform additional rescoring. The best system achieves a 5% relative improvement on an English test-set of real far-field recordings and a 11.6% WER reduction on music-domain utterances.

Jinyu Li, Guoli Ye, Rui Zhao et al.

Recently, the acoustic-to-word model based on the Connectionist Temporal Classification (CTC) criterion was shown as a natural end-to-end model directly targeting words as output units. However, this type of word-based CTC model suffers from the out-of-vocabulary (OOV) issue as it can only model limited number of words in the output layer and maps all the remaining words into an OOV output node. Therefore, such word-based CTC model can only recognize the frequent words modeled by the network output nodes. It also cannot easily handle the hot-words which emerge after the model is trained. In this study, we improve the acoustic-to-word model with a hybrid CTC model which can predict both words and characters at the same time. With a shared-hidden-layer structure and modular design, the alignments of words generated from the word-based CTC and the character-based CTC are synchronized. Whenever the acoustic-to-word model emits an OOV token, we back off that OOV segment to the word output generated from the character-based CTC, hence solving the OOV or hot-words issue. Evaluated on a Microsoft Cortana voice assistant task, the proposed model can reduce the errors introduced by the OOV output token in the acoustic-to-word model by 30%.

Andreas Stolcke, Jasha Droppo

Recent work in automatic recognition of conversational telephone speech (CTS) has achieved accuracy levels comparable to human transcribers, although there is some debate how to precisely quantify human performance on this task, using the NIST 2000 CTS evaluation set. This raises the question what systematic differences, if any, may be found differentiating human from machine transcription errors. In this paper we approach this question by comparing the output of our most accurate CTS recognition system to that of a standard speech transcription vendor pipeline. We find that the most frequent substitution, deletion and insertion error types of both outputs show a high degree of overlap. The only notable exception is that the automatic recognizer tends to confuse filled pauses ("uh") and backchannel acknowledgments ("uhhuh"). Humans tend not to make this error, presumably due to the distinctive and opposing pragmatic functions attached to these words. Furthermore, we quantify the correlation between human and machine errors at the speaker level, and investigate the effect of speaker overlap between training and test data. Finally, we report on an informal "Turing test" asking humans to discriminate between automatic and human transcription error cases.

Zhehuai Chen, Jasha Droppo, Jinyu Li et al.

Unsupervised single-channel overlapped speech recognition is one of the hardest problems in automatic speech recognition (ASR). Permutation invariant training (PIT) is a state of the art model-based approach, which applies a single neural network to solve this single-input, multiple-output modeling problem. We propose to advance the current state of the art by imposing a modular structure on the neural network, applying a progressive pretraining regimen, and improving the objective function with transfer learning and a discriminative training criterion. The modular structure splits the problem into three sub-tasks: frame-wise interpreting, utterance-level speaker tracing, and speech recognition. The pretraining regimen uses these modules to solve progressively harder tasks. Transfer learning leverages parallel clean speech to improve the training targets for the network. Our discriminative training formulation is a modification of standard formulations, that also penalizes competing outputs of the system. Experiments are conducted on the artificial overlapped Switchboard and hub5e-swb dataset. The proposed framework achieves over 30% relative improvement of WER over both a strong jointly trained system, PIT for ASR, and a separately optimized system, PIT for speech separation with clean speech ASR model. The improvement comes from better model generalization, training efficiency and the sequence level linguistic knowledge integration.

Tianxing He, Yu Zhang, Jasha Droppo et al.

We propose to train bi-directional neural network language model(NNLM) with noise contrastive estimation(NCE). Experiments are conducted on a rescore task on the PTB data set. It is shown that NCE-trained bi-directional NNLM outperformed the one trained by conventional maximum likelihood training. But still(regretfully), it did not out-perform the baseline uni-directional NNLM.