Search

Large scale machine learning (ML) systems such as the Alexa automatic speech recognition (ASR) system continue to improve with increasing amounts of manually transcribed training data. Instead of scaling manual transcription to impractical levels, we utilize semi-supervised learning (SSL) to learn acoustic models (AM) from the vast firehose of untranscribed audio data. Learning an AM from 1 Million hours of audio presents unique ML and system design challenges. We present the design and evaluation of a highly scalable and resource efficient SSL system for AM. Employing the student/teacher learning paradigm, we focus on the student learning subsystem: a scalable and robust data pipeline that generates features and targets from raw audio, and an efficient model pipeline, including the distributed trainer, that builds a student model. Our evaluations show that, even without extensive hyper-parameter tuning, we obtain relative accuracy improvements in the 10 to 20$\%$ range, with higher gains in noisier conditions. The end-to-end processing time of this SSL system was 12 days, and several components in this system can trivially scale linearly with more compute resources., 2156-3357 \copyright 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information

The use of spatial information with multiple microphones can improve far-field automatic speech recognition (ASR) accuracy. However, conventional microphone array techniques degrade speech enhancement performance when there is an array geometry mismatch between design and test conditions. Moreover, such speech enhancement techniques do not always yield ASR accuracy improvement due to the difference between speech enhancement and ASR optimization objectives. In this work, we propose to unify an acoustic model framework by optimizing spatial filtering and long short-term memory (LSTM) layers from multi-channel (MC) input. Our acoustic model subsumes beamformers with multiple types of array geometry. In contrast to deep clustering methods that treat a neural network as a black box tool, the network encoding the spatial filters can process streaming audio data in real time without the accumulation of target signal statistics. We demonstrate the effectiveness of such MC neural networks through ASR experiments on the real-world far-field data. We show that our two-channel acoustic model can on average reduce word error rates (WERs) by~13.4 and~12.7% compared to a single channel ASR system with the log-mel filter bank energy (LFBE) feature under the matched and mismatched microphone placement conditions, respectively. Our result also shows that our two-channel network achieves a relative WER reduction of over~7.0% compared to conventional beamforming with seven microphones overall., ICASSP2019, 5 pages. arXiv admin note: substantial text overlap with arXiv:1903.05299

This is a report of our lessons learned building acoustic models from 1 Million hours of unlabeled speech, while labeled speech is restricted to 7,000 hours. We employ student/teacher training on unlabeled data, helping scale out target generation in comparison to confidence model based methods, which require a decoder and a confidence model. To optimize storage and to parallelize target generation, we store high valued logits from the teacher model. Introducing the notion of scheduled learning, we interleave learning on unlabeled and labeled data. To scale distributed training across a large number of GPUs, we use BMUF with 64 GPUs, while performing sequence training only on labeled data with gradient threshold compression SGD using 16 GPUs. Our experiments show that extremely large amounts of data are indeed useful; with little hyper-parameter tuning, we obtain relative WER improvements in the 10 to 20% range, with higher gains in noisier conditions., Comment: "Copyright 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works."

Conventional far-field automatic speech recognition (ASR) systems typically employ microphone array techniques for speech enhancement in order to improve robustness against noise or reverberation. However, such speech enhancement techniques do not always yield ASR accuracy improvement because the optimization criterion for speech enhancement is not directly relevant to the ASR objective. In this work, we develop new acoustic modeling techniques that optimize spatial filtering and long short-term memory (LSTM) layers from multi-channel (MC) input based on an ASR criterion directly. In contrast to conventional methods, we incorporate array processing knowledge into the acoustic model. Moreover, we initialize the network with beamformers' coefficients. We investigate effects of such MC neural networks through ASR experiments on the real-world far-field data where users are interacting with an ASR system in uncontrolled acoustic environments. We show that our MC acoustic model can reduce a word error rate (WER) by~16.5\% compared to a single channel ASR system with the traditional log-mel filter bank energy (LFBE) feature on average. Our result also shows that our network with the spatial filtering layer on two-channel input achieves a relative WER reduction of~9.5\% compared to conventional beamforming with seven microphones., Comment: ICASSP 2019, 5 pages

Statistical TTS systems that directly predict the speech waveform have recently reported improvements in synthesis quality. This investigation evaluates Amazon's statistical speech waveform synthesis (SSWS) system. An in-depth evaluation of SSWS is conducted across a number of domains to better understand the consistency in quality. The results of this evaluation are validated by repeating the procedure on a separate group of testers. Finally, an analysis of the nature of speech errors of SSWS compared to hybrid unit selection synthesis is conducted to identify the strengths and weaknesses of SSWS. Having a deeper insight into SSWS allows us to better define the focus of future work to improve this new technology.

This paper presents our work on building a small-footprint keyword spotting system for a resource-limited language, which requires low CPU, memory and latency. Our keyword spotting system consists of deep neural network (DNN) and hidden Markov model (HMM), which is a hybrid DNN-HMM decoder. We investigate different transfer learning techniques to leverage knowledge and data from a resource-abundant source language to improve the keyword DNN training for a target language which has limited in-domain data. The approaches employed in this paper include training a DNN using source language data to initialize the target language DNN training, mixing data from source and target languages together in a multi-task DNN training setup, using logits computed from a DNN trained on the source language data to regularize the keyword DNN training in the target language, as well as combinations of these techniques. Given different amounts of target language training data, our experimental results show that these transfer learning techniques successfully improve keyword spotting performance for the target language, measured by the area under the curve (AUC) of DNN-HMM decoding detection error tradeoff (DET) curves using a large in-house far-field test set.

We propose a max-pooling based loss function for training Long Short-Term Memory (LSTM) networks for small-footprint keyword spotting (KWS), with low CPU, memory, and latency requirements. The max-pooling loss training can be further guided by initializing with a cross-entropy loss trained network. A posterior smoothing based evaluation approach is employed to measure keyword spotting performance. Our experimental results show that LSTM models trained using cross-entropy loss or max-pooling loss outperform a cross-entropy loss trained baseline feed-forward Deep Neural Network (DNN). In addition, max-pooling loss trained LSTM with randomly initialized network performs better compared to cross-entropy loss trained LSTM. Finally, the max-pooling loss trained LSTM initialized with a cross-entropy pre-trained network shows the best performance, which yields $67.6\%$ relative reduction compared to baseline feed-forward DNN in Area Under the Curve (AUC) measure.

In the past, conventional i-vectors based on a Universal Background Model (UBM) have been successfully used as input features to adapt a Deep Neural Network (DNN) Acoustic Model (AM) for Automatic Speech Recognition (ASR). In contrast, this paper introduces Hidden Markov Model (HMM) based ivectors that use HMM state alignment information from an ASR system for estimating i-vectors. Further, we propose passing these HMM based i-vectors though an explicit non-linear hidden layer of a DNN before combining them with standard acoustic features, such as log filter bank energies (LFBEs). To improve robustness to mismatched adaptation data, we also propose estimating i-vectors in a causal fashion for training the DNN, restricting the connectivity among hidden nodes in the DNN and applying a max-pool non-linearity at selected hidden nodes. In our experiments, these techniques yield about 5-7% relative word error rate (WER) improvement over the baseline speaker independent system in matched condition, and a substantial WER reduction for mismatched adaptation data.

We investigate the problem of speaker adaptation of DNN acoustic models in two settings: the traditional unsupervised adaptation and a supervised adaptation (SuA) where a few minutes of transcribed speech is available. SuA presents additional difficulties when a test speaker’s adaptation information does not match the registered speaker’s information. Employing feature-space maximum likelihood linear regression (fMLLR) transformed features as side-information to the DNN, we reintroduce some classical ideas for combining adapted and unadapted features: early and late fusion methods, as well as the estimation of the fMLLR transforms using simple target models (STM). Results show that early fusion helps DNNs generalize better when features are combined after a non-linear bottleneck layer, while late fusion improves robustness, specifically in mismatched cases. STM give consistent improvements in both settings.

Described are methods and systems for reducing the audible gap in concatenated recorded speech, resulting in more natural sounding speech in voice applications. The sound of concatenated, recorded speech is improved by also coarticulating the recorded speech. The resulting message is smooth, natural sounding and lifelike. Existing libraries of regularly recorded bulk prompts can be used by coarticulating the user interface prompt occurring just before the bulk prompt. Applications include phone-based applications as well as non-phone-based applications.