In the era of large models, the autoregressive nature of decoding often results in latency serving as a significant bottleneck. We propose a non-autoregressive LM-fused ASR system that effectively leverages the parallelization capabilities of accelerator hardware. Our approach combines the Universal Speech Model (USM) and the PaLM 2 language model in per-segment scoring mode, achieving an average relative WER improvement across all languages of 10.8% on FLEURS and 3.6% on YouTube captioning. Furthermore, our comprehensive ablation study analyzes key parameters such as LLM size, context length, vocabulary size, fusion methodology. For instance, we explore the impact of LLM size ranging from 128M to 340B parameters on ASR performance. This study provides valuable insights into the factors influencing the effectiveness of practical large-scale LM-fused speech recognition systems.

We propose Easy End-to-End Diffusion-based Text to Speech, a simple and efficient end-to-end text-to-speech model based on diffusion. E3 TTS directly takes plain text as input and generates an audio waveform through an iterative refinement process. Unlike many prior work, E3 TTS does not rely on any intermediate representations like spectrogram features or alignment information. Instead, E3 TTS models the temporal structure of the waveform through the diffusion process. Without relying on additional conditioning information, E3 TTS could support flexible latent structure within the given audio. This enables E3 TTS to be easily adapted for zero-shot tasks such as editing without any additional training. Experiments show that E3 TTS can generate high-fidelity audio, approaching the performance of a state-of-the-art neural TTS system. Audio samples are available at https://e3tts.github.io.

We investigate using large language models (LLMs) to generate text-only training data for improving multilingual and code-switching automatic speech recognition (ASR) through a text injection method. In a multilingual setup or a low-resource scenario such as code-switching, we propose to generate text data using the state-of-the-art PaLM 2. To better match the generated text data with specific tasks, we use prompt tuning to adapt PaLM 2 to generate domain-relevant multilingual or code-switched text data for text injection. We can achieve significant improvements in Word Error Rate (WER) in both multilingual and code-switching scenarios. The multilingual experiment shows a $6.2 \%$ relative WER reduction on average, i.e., from $11.25 \%$ to $10.55 \%$, compared to a baseline without text injection. The improvement is up to $23.1 \%$ improvement for certain languages. While in the code-switching scenario, we use English-only prompts to generate Mandarin-English code-switching text and achieve a 3.6% relative WER reduction for a code-switching test set, as well as WER reductions in both English and Mandarin monolingual scenarios, $5.3 \%$ and $8.5 \%$ relative, respectively. Our findings demonstrate that leveraging LLMs for text generation and then injection benefits multilingual or code-switching ASR tasks.

We present a joint Speech and Language Model (SLM), a multitask, multilingual, and dual-modal model that takes advantage of pretrained foundational speech and language models. SLM freezes the pretrained foundation models to maximally preserves their capabilities, and only trains a simple adapter with just 1% (156M) of the foundation models’ parameters. This adaptation not only leads SLM to achieve strong performance on conventional tasks such as automatic speech recognition (ASR) and automatic speech translation (AST), but also unlocks the novel capability of zero-shot instruction-following for more diverse tasks. Given a speech input and a text instruction, SLM is able to perform unseen generation tasks including contextual biasing ASR using real-time context, dialog generation, speech continuation, and question answering. Our approach demonstrates that the representational gap between pretrained speech and language models is narrower than one would expect, and can be bridged by a simple adaptation mechanism. As a result, SLM is not only efficient to train, but also inherits strong capabilities already present in foundation models of different modalities.

While large language models (LLM) have made impressive progress in natural language processing, it remains unclear how to utilize them in improving automatic speech recognition (ASR). In this work, we propose to train a single multilingual language model (LM) for shallow fusion in multiple languages. We push the limits of the multilingual LM to cover up to 84 languages by scaling up using a mixture-of-experts LLM, i.e., generalist language model (GLaM). When the number of experts increases, GLaM dynamically selects only two at each decoding step to keep the inference computation roughly constant. We then apply GLaM to a multilingual shallow fusion task based on a state-of-the-art end-to-end model. Compared to a dense LM of similar computation during inference, GLaM reduces the WER of an English long-tail test set by 4.4% relative. In a multilingual shallow fusion task, GLaM improves 41 out of 50 languages with an average relative WER reduction of 3.85%, and a maximum reduction of 10%. Compared to the baseline model, GLaM achieves an average WER reduction of 5.53% over 43 languages.

Knowledge distillation is an effective machine learning technique to transfer knowledge from a teacher model to a smaller student model, especially with unlabeled data. In this paper, we focus on knowledge distillation for the RNN-T model, which is widely used in state-of-the-art (SoTA) automatic speech recognition (ASR). Specifically, we compared using soft and hard target distillation to train large-scale RNN-T models on the LibriSpeech/LibriLight public dataset (60k hours) and our in-house data (600k hours). We found that hard targets are more effective when the teacher and student have different architecture, such as large teacher and small streaming student. On the other hand, soft target distillation works better in self-training scenario like iterative large teacher training. For a large model with 0.6B weights, we achieve a new SoTA word error rate (WER) on LibriSpeech (8% relative improvement on dev-other) using Noisy Student Training with soft target distillation. It also allows our production teacher to adapt new data domain continuously.

Foundation models (FMs), that are trained on broad data at scale and are adaptable to a wide range of downstream tasks, have brought large interest in the research community. Benefiting from the diverse data sources such as different modalities, languages and application domains, foundation models have demonstrated strong generalization and knowledge transfer capabilities. In this paper, we present a pioneering study towards building an efficient solution for FM-based speech recognition systems. We adopt the recently developed self-supervised BEST-RQ for pretraining, and extend the joint training strategy JUST Hydra for finetuning using both source and unsuper-vised target domain data. The FM encoder adapter and decoder are then finetuned to the target domain with a small amount of super-vised in-domain data. On a large-scale YouTube and Voice Search task, our method is shown to be both data and model parameter efficient. It achieves the same quality with only 21.6M supervised in-domain data and 130.8M finetuned parameters, compared to the 731.1M model trained from scratch on additional 300M supervised in-domain data.

We propose JEIT, a joint end-to-end (E2E) model and internal language model (ILM) training method to inject large-scale unpaired text into ILM during E2E training which improves rare-word speech recognition. With JEIT, the E2E model computes an E2E loss on audio-transcript pairs while its ILM estimates a cross-entropy loss on unpaired text. The E2E model is trained to minimize a weighted sum of E2E and ILM losses. During JEIT, ILM absorbs knowledge from unpaired text while the E2E training serves as regularization. Unlike ILM adaptation methods, JEIT does not require a separate adaptation step and avoids the need for Kullback-Leibler divergence regularization of ILM. We also show that modular hybrid autoregressive transducer (MHAT) performs better than HAT in the JEIT framework, and is much more robust than HAT during ILM adaptation. To push the limit of unpaired text injection, we further propose a combined JEIT and JOIST training (CJJT) that benefits from modality matching, encoder text injection and ILM training. Both JEIT and CJJT can foster a more effective LM fusion. With 100B unpaired sentences, JEIT/CJJT improves rare-word recognition accuracy by up to 16.4% over a model trained without unpaired text.

This paper proposes Virtuoso, a massively multilingual speech–text joint semi-supervised learning framework for text-to-speech synthesis (TTS) models. Existing multilingual TTS typically supports tens of languages, which are a small fraction of the thousands of languages in the world. One difficulty to scale multilingual TTS to hundreds of languages is collecting high-quality speech–text paired data in low-resource languages. This study extends Maestro, a speech–text joint pretraining framework for automatic speech recognition (ASR), to speech generation tasks. To train a TTS model from various types of speech and text data, different training schemes are designed to handle supervised (paired TTS and ASR data) and unsupervised (untranscribed speech and unspoken text) datasets. Experimental evaluation shows that 1) multilingual TTS models trained on Virtuoso can achieve significantly better naturalness and intelligibility than baseline ones in seen languages, and 2) they can synthesize reasonably intelligible and naturally sounding speech for unseen languages where no high-quality paired TTS data is available.

We propose a novel method to accelerate training and inference process of recurrent neural network transducer (RNN-T) based on the guidance from a co-trained connectionist temporal classification (CTC) model. We made a key assumption that if an encoder embedding frame is classified as a blank frame by the CTC model, it is likely that this frame will be aligned to blank for all the partial alignments or hypotheses in RNN-T and it can be discarded from the decoder input. We also show that this frame reduction operation can be applied in the middle of the encoder, which result in significant speed up for the training and inference in RNN-T. We further show that the CTC alignment, a by-product of the CTC decoder, can also be used to perform lattice reduction for RNN-T during training. Our method is evaluated on the Librispeech and SpeechStew tasks. We demonstrate that the proposed method is able to accelerate the RNN-T inference by 2.2 times with similar or slightly better word error rates (WER).

Recently, a number of approaches to train speech models by incorporating text into end-to-end models have been developed, with Maestro advancing state-of-the-art automatic speech recognition (ASR) and Speech Translation (ST) performance. In this paper, we expand our understanding of the resulting shared speech-text representations with two types of analyses. First we examine the limits of speech-free domain adaptation, finding that a corpus-specific duration model for speech-text alignment is the most important component for learning a shared speech-text representation. Second, we inspect the similarities between activations of unimodal (speech or text) encoders as compared to the activations of a shared encoder. We find that the shared encoder learns a more compact and overlapping speech-text representation than the uni-modal encoders. We hypothesize that this partially explains the effectiveness of the Maestro shared speech-text representations.

In this work, we propose a new parameter-efficient learning framework based on neural model reprogramming for cross-lingual speech recognition, which can re-purpose well-trained English automatic speech recognition (ASR) models to recognize the other languages. We design different auxiliary neural architectures focusing on learnable pre-trained feature enhancement that, for the first time, empowers model reprogramming on ASR. Specifically, we investigate how to select trainable components (i.e., encoder) of a conformer-based RNN-Transducer, as a frozen pre-trained backbone. Experiments on a seven-language multilingual LibriSpeech speech (MLS) task show that model reprogramming only requires 4.2% (11M out of 270M) to 6.8% (45M out of 660M) of its original trainable parameters from a full ASR model to perform competitive results in a range of 11.9% to 8.1% WER averaged across different languages. In addition, we discover different setups to make large-scale pre-trained ASR succeed in both monolingual and multilingual speech recognition. Our methods outperform existing ASR tuning architectures and their extension with self-supervised losses (e.g., w2v-bert) in terms of lower WER and better training efficiency.

We propose a quantum kernel learning (QKL) framework to address the inherent data sparsity issues often encountered in training large-scare acoustic models in low-resource scenarios. We project acoustic features based on classical-to-quantum feature encoding. Different from existing quantum convolution techniques, we utilize QKL with features in the quantum space to design kernel-based classifiers. Experimental results on challenging spoken command recognition tasks for a few low-resource languages, such as Arabic, Georgian, Chuvash, and Lithuanian, show that the proposed QKL-based hybrid approach attains good improvements over existing classical and quantum solutions.

We present Mu$^2$SLAM, a multilingual sequence-to-sequence model pre-trained jointly on unlabeled speech, unlabeled text and supervised data spanning Automatic Speech Recognition (ASR), Automatic Speech Translation (AST) and Machine Translation (MT), in over 100 languages. By leveraging a quantized representation of speech as a target, Mu$^2$SLAM trains the speech-text models with a sequence-to-sequence masked denoising objective similar to T5 on the decoder and a masked language modeling objective (MLM) on the encoder, for both unlabeled speech and text, while utilizing the supervised tasks to improve cross-lingual and cross-modal representation alignment within the model. On CoVoST AST, Mu$^2$SLAM establishes a new state-of-the-art for models trained on public datasets, improving on xx-en translation over the previous best by 1.9 BLEU points and on en-xx translation by 1.1 BLEU points. On Voxpopuli ASR, our model matches the performance of an mSLAM model fine-tuned with an RNN-T decoder, despite using a relatively weaker Transformer decoder. On text understanding tasks, our model improves by more than 6% over mSLAM on XNLI, getting closer to the performance of mT5 models of comparable capacity on XNLI and TydiQA, paving the way towards a single model for all speech and text understanding tasks.

In this work, we introduce a "score-based assessment" framework for estimating the transferability of pre-trained speech models (PSMs) for fine-tuning target tasks.We leverage upon two representation theories, Bayesian likelihood estimation and optimal transport, to generate rank scores for the PSM candidates using the extracted representations.Our framework efficiently computes transferability scores without actual finetuning of candidate models or layers by making a temporal independent hypothesis.We evaluate some popular supervised speech models (e.g., Conformer RNN-Transducer) and selfsupervised speech models (e.g., HuBERT) in cross-layer and cross-model settings using public data.Experimental results show a high Spearman's rank correlation and low p-value between our estimation framework and fine-tuning ground truth.Our proposed transferability framework requires less computational time and resources, making it a resource-saving and timeefficient approach for tuning speech foundation models.

We propose Conditional Adapter (CoDA), a parameter-efficient transfer learning method that also improves inference efficiency. CoDA generalizes beyond standard adapter approaches to enable a new way of balancing speed and accuracy using conditional computation. Starting with an existing dense pretrained model, CoDA adds sparse activation together with a small number of new parameters and a light-weight training phase. Our experiments demonstrate that the CoDA approach provides an unexpectedly efficient way to transfer knowledge. Across a variety of language, vision, and speech tasks, CoDA achieves a 2x to 8x inference speed-up compared to the state-of-the-art Adapter approaches with moderate to no accuracy loss and the same parameter efficiency.

Self-supervised training has shown promising gains in pretraining models and facilitating the downstream finetuning for speech recognition, like multilingual ASR. Most existing methods adopt a 2-stage scheme where the self-supervised loss is optimized in the first pretraining stage, and the standard supervised finetuning resumes in the second stage. In this paper, we propose an end-to-end (E2E) Joint Unsupervised and Supervised Training (JUST) method to combine the supervised RNN-T loss and the self-supervised contrastive and masked language modeling (MLM) losses. We validate its performance on the public dataset Multilingual LibriSpeech (MLS), which includes 8 languages and is extremely imbalanced. On MLS, we explore (1) JUST trained from scratch, and (2) JUST finetuned from a pretrained checkpoint. Experiments show that JUST can consistently outperform other existing state-of-the-art methods, and beat the monolingual baseline by a significant margin, demonstrating JUST’s capability of handling low-resource languages in multilingual ASR. Our average WER of all languages outperforms average monolingual baseline by 33.3%, and the state-of-the-art 2-stage XLSR by 32%. On low-resource languages like Polish, our WER is less than half of the monolingual baseline and even beats the supervised transfer learning method which uses external supervision.

An effective way to learn representations from untranscribed speech and unspoken text with linguistic/lexical representations derived from synthesized speech was introduced in tts4pretrain [1]. However, the representations learned from synthesized and real speech are likely to be different, potentially limiting the improvements from incorporating unspoken text. In this paper, we introduce learning from supervised speech earlier on in the training process with consistency-based regularization between real and synthesized speech. This allows for better learning of shared speech and text representations. Thus, we introduce a new objective, with encoder and decoder consistency and contrastive regularization between real and synthesized speech derived from the labeled corpora during the pretraining stage. We show that the new objective leads to more similar representations derived from speech and text that help downstream ASR. The proposed pretraining method yields Word Error Rate (WER) reductions of 7-21% relative on six public corpora, Librispeech, AMI, TEDLIUM, Common Voice, Switchboard, CHiME-6, over a state-of-the-art baseline pretrained with wav2vec2.0 and 2-17% over the previously proposed tts4pretrain. The proposed method outperforms the supervised SpeechStew by up to 17%. Moreover, we show that the proposed method also yields WER reductions on larger data sets by evaluating on a large resource, in-house Voice Search task and streaming ASR.

The development of end-to-end models has largely sped up the research in massively multilingual automatic speech recognition (MMASR). Previous research has demonstrated the feasibility to build high quality MMASR models. In this work, we study the impact of adding more languages and propose a lifelong learning approach to build high quality MMASR systems. Experiments on a 66-language Voice Search task show that we can take a model built on 15 languages and continue training to obtain a 32-language model and similarly to further build a 67-language model. More importantly, models developed in this way achieve better quality compared to those trained from scratch. It maintains similar performance on old languages and achieves competitive results on new ones. This would potentially speed up the development of universal ASR models that recognize speech from any language, any domain and any environment by reusing knowledge learned beforehand.

As end-to-end automatic speech recognition (ASR) models reach promising performance, various downstream tasks rely on good confidence estimators for these systems. Recent research has shown that model-based confidence estimators have a significant advantage over using the output softmax probabilities. If the input data to the speech recogniser is from mismatched acoustic and linguistic conditions, the ASR performance and the corresponding confidence estimators may exhibit severe degradation. Since confidence models are often trained on the same in-domain data as the ASR, generalising to out-of-domain (OOD) scenarios is challenging. By keeping the ASR model untouched, this paper proposes two approaches to improve the model-based confidence estimators on OOD data: using pseudo transcriptions and an additional OOD language model. With an ASR model trained on LibriSpeech, experiments show that the proposed methods can greatly improve the confidence metrics on TED-LIUM and Switchboard datasets while preserving in-domain performance. Furthermore, the improved confidence estimators are better calibrated on OOD data and can provide a much more reliable criterion for data selection.

In this paper, we explore reducing computational latency of the 2-pass cascaded encoder model [1]. Specifically, we experiment with reducing the size of the causal 1st-pass and adding capacity to the non-causal 2nd-pass, such that the overall latency can be reduced without loss of quality. In addition, we explore using a confidence model for deciding to stop 2nd-pass recognition if we are confident in the 1st-pass hypothesis. Overall, we are able to reduce latency by a factor of 1.7X, compared to the baseline cascaded encoder from [1]. Secondly, with the added capacity in the non-causal 2nd-pass, we find that we can improve WER by up to 7% relative using wav2vec and minimum word-error-rate (MWER) training.

Many speech applications require understanding aspects beyond the words being spoken, such as recognizing emotion, detecting whether the speaker is wearing a mask, or distinguishing real from synthetic speech. In this work, we introduce a new state-of-the-art paralinguistic representation derived from large-scale, fully self-supervised training of a 600M+ parameter Conformer-based architecture. We benchmark on a diverse set of speech tasks and demonstrate that simple linear classifiers trained on top of our time-averaged representation outperform nearly all previous results, in some cases by large margins. Our analyses of context-window size demonstrate that, surprisingly, 2 second context-windows achieve 96% the performance of the Conformers that use the full long-term context on 7 out of 9 tasks. Furthermore, while the best per-task representations are extracted internally in the network, stable performance across several layers allows a single universal representation to reach near optimal performance on all tasks.

We present Maestro, a self-supervised training method to unify representations learnt from speech and text modalities.Self-supervised learning from speech signals aims to learn the latent structure inherent in the signal, while self-supervised learning from text attempts to capture lexical information.Learning aligned representations from unpaired speech and text sequences is a challenging task.Previous work either implicitly enforced the representations learnt from these two modalities to be aligned in the latent space through multitasking and parameter sharing or explicitly through conversion of modalities via speech synthesis.While the former suffers from interference between the two modalities, the latter introduces additional complexity.In this paper, we propose Maestro, a novel algorithm to learn unified representations from both these modalities simultaneously that can transfer to diverse downstream tasks such as Automated Speech Recognition (ASR) and Speech Translation (ST).Maestro learns unified representations through sequence alignment, duration prediction and matching embeddings in the learned space through an aligned masked-language model loss.We establish a new state-of-the-art (SOTA) on VoxPopuli multilingual ASR with a 8% relative reduction in Word Error Rate (WER), multidomain SpeechStew ASR (3.7% relative) and 21 languages to English multilingual ST on CoVoST 2 with an improvement of 2.8 BLEU averaged over 21 languages.

We introduce XTREME-S, a new benchmark to evaluate universal cross-lingual speech representations in many languages.XTREME-S covers four task families: speech recognition, classification, speech-to-text translation and retrieval.Covering 102 languages from 10+ language families, 3 different domains and 4 task families, XTREME-S aims to simplify multilingual speech representation evaluation, as well as catalyze research in "universal" speech representation learning.This paper describes the new benchmark and establishes the first speech-only and speechtext baselines using XLS-R and mSLAM on all downstream tasks.We motivate the design choices and detail how to use the benchmark.Datasets and fine-tuning scripts are made easily accessible through the HuggingFace platform. 1

Transfer tasks in text-to-speech (TTS) synthesis -where one or more aspects of the speech of one set of speakers is transferred to another set of speakers that do not feature these aspects originally -remains a challenging task.One of the challenges is that models that have high-quality transfer capabilities can have issues in stability, making them impractical for user-facing critical tasks.This paper demonstrates that transfer can be obtained by training a robust TTS system on data generated by a less robust TTS system designed for a high-quality transfer task; in particular, a CHiVE-BERT monolingual TTS system is trained on the output of a Tacotron model designed for accent transfer.While some quality loss is inevitable with this approach, experimental results show that the models trained on synthetic data this way can produce high quality audio displaying accent transfer, while preserving speaker characteristics such as speaking style.

End-to-end speech-to-speech translation (S2ST) without relying on intermediate text representations is a rapidly emerging frontier of research.Recent works have demonstrated that the performance of such direct S2ST systems is approaching that of conventional cascade S2ST when trained on comparable datasets.However, in practice, the performance of direct S2ST is bounded by the availability of paired S2ST training data.In this work, we explore multiple approaches for leveraging much more widely available unsupervised and weakly-supervised speech and text data to improve the performance of direct S2ST based on Translatotron 2. With our most effective approaches, the average translation quality of direct S2ST on 21 language pairs on the CVSS-C corpus is improved by +13.6 BLEU (or +113% relatively), as compared to the previous state-of-the-art trained without additional data.The improvements on low-resource language are even more significant (+398% relatively on average).Our comparative studies suggest future research directions for S2ST and speech representation learning.

Self-supervised learning of speech representations has achieved impressive results in improving automatic speech recognition (ASR).In this paper, we show that data selection is important for self-supervised learning.We propose a simple and effective unsupervised data selection method which selects acoustically similar speech to a target domain.It takes the discrete speech representation available in common self-supervised learning frameworks as input, and applies a contrastive data selection method on the discrete tokens.Through extensive empirical studies we show that our proposed method reduces the amount of required pre-training data and improves the downstream ASR performance.Pre-training on a selected subset of 6% of the general data pool results in 11.8% relative improvements in LibriSpeech test-other compared to pre-training on the full set.On Multilingual LibriSpeech French, German, and Spanish test sets, selecting 6% data for pre-training reduces word error rate by more than 15% relatively compared to the full set, and achieves competitive results compared to current state-of-theart performances.

Training state-of-the-art Automated Speech Recognition (ASR) models typically requires a substantial amount of transcribed speech. In this work, we demonstrate that a modality-matched joint speech and text model introduced in [1] can be leveraged to train a massively multilingual ASR model without any supervised (manually transcribed) speech for some languages. This paper explores the use of jointly learnt speech and text representations in a massively multilingual, zero supervised speech, real-world setting to expand the set of languages covered by ASR with only unlabeled speech and text in the target languages. Using the FLEURS dataset, we define the task to cover 102 languages, where transcribed speech is available in 52 of these languages and can be used to improve end-to-end ASR quality on the remaining 50. First, we show that by combining speech representations with byte-level text representations and use of language embeddings, we can dramatically reduce the Character Error Rate (CER) on languages with no supervised speech from 64.8% to 30.8%, a relative reduction of 53%. Second, using a subset of South Asian languages we show that Maestro-U can promote knowledge transfer from languages with supervised speech even when there is limited to no graphemic overlap. Overall, Maestro-U closes the gap to oracle performance by 68.5% relative and reduces the CER of 19 languages below 15%.

We introduce FLEURS, the Few-shot Learning Evaluation of Universal Representations of Speech benchmark. FLEURS is an n-way parallel speech dataset in 102 languages built on top of the machine translation FLoRes-101 benchmark, with approximately 12 hours of speech supervision per language. FLEURS can be used for a variety of speech tasks, including Automatic Speech Recognition (ASR), Speech Language Identification (Speech LangID), Speech-Text Retrieval. In this paper, we provide baselines for the tasks based on multilingual pre-trained models like speech-only w2v-BERT [1] and speech-text multimodal mSLAM [2]. The goal of FLEURS is to enable speech technology in more languages and catalyze research in low-resource speech understanding.1.

Self-supervised learned (SSL) speech pre-trained models perform well across various speech processing tasks. Distilled versions of SSL models have been developed to match the needs of on-device speech applications. Though having similar performance as original SSL models, distilled counterparts suffer from performance degradation even more than their original versions in distorted environments. This paper proposes to apply Cross-Distortion Mapping and Domain Adversarial Training to SSL models during knowledge distillation to alleviate the performance gap caused by the domain mismatch problem. Results show consistent performance improvements under both in- and out-of-domain distorted setups for different downstream tasks while keeping efficient model size.

Text-only adaptation of a transducer model remains challenging for end-to-end speech recognition since the transducer has no clearly separated acoustic model (AM), language model (LM) or blank model. In this work, we propose a modular hybrid autoregressive transducer (MHAT) that has structurally separated label and blank decoders to predict label and blank distributions, respectively, along with a shared acoustic encoder. The encoder and label decoder outputs are directly projected to AM and internal LM scores and then added to compute label posteriors. We train MHAT with an internal LM loss and a HAT loss to ensure that its internal LM becomes a standalone neural LM that can be effectively adapted to text. Moreover, text adaptation of MHAT fosters a much better LM fusion than internal LM subtraction-based methods. On Google's large-scale production data, a multi-domain MHAT adapted with 100B sentences achieves relative WER reductions of up to 12.4% without LM fusion and 21.5% with LM fusion from 400K-hour trained HAT.

We present JOIST, an algorithm to train a streaming, cascaded, encoder end-to-end (E2E) model with both speech-text paired inputs, and text-only unpaired inputs. Unlike previous works, we explore joint training with both modalities, rather than pre-training and fine-tuning. In addition, we explore JOIST using a streaming E2E model with an order of magnitude more data, which are also novelties compared to previous works. Through a series of ablation studies, we explore different types of text modeling, including how to model the length of the text sequence and the appropriate text subword unit representation. We find that best text representation for JOIST improves WER across a variety of search and rare-word test sets by 4-14% relative, compared to a model not trained with text. In addition, we quantitatively show that JOIST maintains streaming capabilities, which is important for good user-level experience.

Self-supervised pretraining for Automated Speech Recognition (ASR) has shown varied degrees of success. In this paper, we propose to jointly learn representations during pretraining from two different modalities: speech and text. The proposed method, tts4pretrain complements the power of contrastive learning in self-supervision with linguistic/lexical representations derived from synthesized speech, effectively learning from untranscribed speech and unspoken text. Lexical learning in the speech encoder is enforced through an additional sequence loss term that is coupled with contrastive loss during pretraining. We demonstrate that this novel pretraining method yields Word Error Rate (WER) reductions of 10% relative on the well-benchmarked, Librispeech task over a state-of-the-art baseline pretrained with wav2vec2.0 only. The proposed method also serves as an effective strategy to compensate for the lack of transcribed speech, effectively matching the performance of 5000 hours of transcribed speech with just 100 hours of transcribed speech on the AMI meeting transcription task. Finally, we demonstrate WER reductions of up to 15% on an inhouse Voice Search task over traditional pretraining. Incorporating text into encoder pretraining is complimentary to rescoring with a larger or in-domain language model, resulting in additional 6% relative reduction in WER.

Motivated by the success of masked language modeling (MLM) in pre-training natural language processing models, we propose w2v-BERT that explores MLM for self-supervised speech representation learning. w2v-BERT is a framework that combines contrastive learning and MLM, where the former trains the model to discretize input continuous speech signals into a finite set of discriminative speech tokens, and the latter trains the model to learn contextualized speech representations via solving a masked prediction task consuming the discretized tokens. In contrast to existing MLM-based speech pre-training frameworks such as HuBERT, which relies on an iterative re-clustering and re-training process, or vq-wav2vec, which concatenates two separately trained modules, w2v-BERT can be optimized in an end-to-end fashion by solving the two self-supervised tasks (the contrastive task and MLM) simultaneously. Our experiments show that w2v-BERT achieves competitive results compared to current state-of-the-art pre-trained models on the LibriSpeech benchmarks when using the Libri-Light 60k corpus as the unsupervised data. In particular, when compared to published models such as conformer-based wav2vec 2.0 and HuBERT, our model shows 5% to 10% relative WER reduction on the test-clean and test-other subsets. When applied to the Google's Voice Search traffic dataset, w2v-BERT outperforms our internal conformer-based wav2vec 2.0 by more than 30% relatively.

Building ASR models across many languages is a challenging multi-task learning problem due to large variations and heavily unbalanced data. Existing work has shown positive transfer from high resource to low resource languages. However, degradations on high resource languages are commonly observed due to interference from the heterogeneous multilingual data and reduction in per-language capacity. We conduct a capacity study on a 15-language task, with the amount of data per language varying from 7.6K to 53.5K hours. We adopt GShard [1] to efficiently scale up to 10B parameters. Empirically, we find that (1) scaling the number of model parameters is an effective way to solve the capacity bottleneck - our 500M-param model already outperforms monolingual baselines and scaling it to 1B and 10B brought further quality gains; (2) larger models are not only more data efficient, but also more efficient in terms of training cost as measured in TPU days - the 1B-param model reaches the same accuracy at 34% of training time as the 500M-param model; (3) given a fixed capacity budget, adding depth works better than width and large encoders do better than large decoders; (4) with continuous training, they can be adapted to new languages and domains.

Streaming end-to-end automatic speech recognition (ASR) models are widely used on smart speakers and on-device applications. Since these models are expected to transcribe speech with minimal latency, they are constrained to be causal with no future context, compared to their non-streaming counterparts. Consequently, streaming models usually perform worse than non-streaming models. We propose a novel and effective learning method by leveraging a non-streaming ASR model as a teacher to generate transcripts on an arbitrarily large data set, which is then used to distill knowledge into streaming ASR models. This way, we scale the training of streaming models to up to 3 million hours of YouTube audio. Experiments show that our approach can significantly reduce the word error rate (WER) of RNN-T models not only on LibriSpeech but also on YouTube data in four languages. For example, in French, we are able to reduce the WER by 16.4% relatively to a baseline streaming model by leveraging a non-streaming teacher model trained on the same amount of labeled data as the baseline.

Although neural end-to-end text-to-speech models can synthesize highly natural speech, there is still room for improvements to its efficiency and naturalness. This paper proposes a non-autoregressive neural text-to-speech model augmented with a variational autoencoder-based residual encoder. This model, called Parallel Tacotron, is highly parallelizable during both training and inference, allowing efficient synthesis on modern parallel hardware. The use of the variational autoencoder relaxes the one-to-many mapping nature of the text-to-speech problem and improves naturalness. To further improve the naturalness, we use lightweight convolutions, which can efficiently capture local contexts, and introduce an iterative spectrogram loss inspired by iterative refinement. Experimental results show that Parallel Tacotron matches a strong autoregressive baseline in subjective evaluations with significantly decreased inference time.

End-to-end (E2E) models have shown to outperform state-of-the-art conventional models for streaming speech recognition [1] across many dimensions, including quality (as measured by word error rate (WER)) and endpointer latency [2]. However, the model still tends to delay the predictions towards the end and thus has much higher partial latency compared to a conventional ASR model. To address this issue, we look at encouraging the E2E model to emit words early, through an algorithm called FastEmit [3]. Naturally, improving on latency results in a quality degradation. To address this, we explore replacing the LSTM layers in the encoder of our E2E model with Conformer layers [4], which has shown good improvements for ASR. Secondly, we also explore running a 2nd-pass beam search to improve quality. In order to ensure the 2nd-pass completes quickly, we explore non-causal Conformer layers that feed into the same 1st-pass RNN-T decoder, an algorithm called Cascaded Encoders [5]. Overall, the Conformer RNN-T with Cascaded Encoders offers a better quality and latency tradeoff for streaming ASR.

For various speech-related tasks, confidence scores from a speech recogniser are a useful measure to assess the quality of transcriptions. In traditional hidden Markov model-based automatic speech recognition (ASR) systems, confidence scores can be reliably obtained from word posteriors in decoding lattices. However, for an ASR system with an auto-regressive decoder, such as an attention-based sequence-to-sequence model, computing word posteriors is difficult. An obvious alternative is to use the decoder softmax probability as the model confidence. In this paper, we first examine how some commonly used regularisation methods influence the softmax-based confidence scores and study the overconfident behaviour of end-to-end models. Then we propose a lightweight and effective approach named confidence estimation module (CEM) on top of an existing end-to-end ASR model. Experiments on LibriSpeech show that CEM can mitigate the overconfidence problem and can produce more reliable confidence scores with and without shallow fusion of a language model. Further analysis shows that CEM generalises well to speech from a moderately mismatched domain and can potentially improve downstream tasks such as semi-supervised learning.

We study the problem of word-level confidence estimation in subword-based end-to-end (E2E) models for automatic speech recognition (ASR). Although prior works have proposed training auxiliary confidence models for ASR systems, they do not extend naturally to systems that operate on word-pieces (WP) as their vocabulary. In particular, ground truth WP correctness labels are needed for training confidence models, but the non-unique tokenization from word to WP causes inaccurate labels to be generated. This paper proposes and studies two confidence models of increasing complexity to solve this problem. The final model uses self-attention to directly learn word-level confidence without needing subword tokenization, and exploits full context features from multiple hypotheses to improve confidence accuracy. Experiments on Voice Search and long-tail test sets show standard metrics (e.g., NCE, AUC, RMSE) improving substantially. The proposed confidence module also enables a model selection approach to combine an on-device E2E model with a hybrid model on the server to address the rare word recognition problem for the E2E model.

We propose automatic speech recognition (ASR) models inspired by echo state network (ESN) [1], in which a subset of recurrent neural networks (RNN) layers in the models are randomly initialized and untrained. Our study focuses on RNN-T and Conformer models, and we show that model quality does not drop even when the decoder is fully randomized. Furthermore, such models can be trained more efficiently as the decoders do not require to be updated. By contrast, randomizing encoders hurts model quality, indicating that optimizing encoders and learn proper representations for acoustic inputs are more vital for speech recognition. Overall, we challenge the common practice of training ASR models for all components, and demonstrate that ESN-based models can perform equally well but enable more efficient training and storage than fully-trainable counterparts.

This paper introduces WaveGrad 2, a non-autoregressive generative model for text-to-speech synthesis. WaveGrad 2 is trained to estimate the gradient of the log conditional density of the waveform given a phoneme sequence. The model takes an input phoneme sequence, and through an iterative refinement process, generates an audio waveform. This contrasts to the original WaveGrad vocoder which conditions on mel-spectrogram features, generated by a separate model. The iterative refinement process starts from Gaussian noise, and through a series of refinement steps (e.g., 50 steps), progressively recovers the audio sequence. WaveGrad 2 offers a natural way to trade-off between inference speed and sample quality, through adjusting the number of refinement steps. Experiments show that the model can generate high fidelity audio, approaching the performance of a state-of-the-art neural TTS system. We also report various ablation studies over different model configurations. Audio samples are available at this https URL.

This paper introduces Parallel Tacotron 2, a non-autoregressive neural text-to-speech model with a fully differentiable duration model which does not require supervised duration signals.The duration model is based on a novel attention mechanism and an iterative reconstruction loss based on Soft Dynamic Time Warping, this model can learn token-frame alignments as well as token durations automatically.Experimental results show that Parallel Tacotron 2 outperforms baselines in subjective naturalness in several diverse multi speaker evaluations.

This paper introduces PnG BERT, a new encoder model for neural TTS.This model is augmented from the original BERT model, by taking both phoneme and grapheme representations of text as input, as well as the word-level alignment between them.It can be pre-trained on a large text corpus in a selfsupervised manner, and fine-tuned in a TTS task.Experimental results show that a neural TTS model using a pre-trained PnG BERT as its encoder yields more natural prosody and more accurate pronunciation than a baseline model using only phoneme input with no pre-training.Subjective side-by-side preference evaluations show that raters have no statistically significant preference between the speech synthesized using a PnG BERT and ground truth recordings from professional speakers.

End-to-end models with auto-regressive decoders have shown impressive results for automatic speech recognition (ASR). These models formulate the sequence-level probability as a product of the conditional probabilities of all individual tokens given their histories. However, the performance of locally normalised models can be sub-optimal because of factors such as exposure bias. Consequently, the model distribution differs from the underlying data distribution. In this paper, the residual energy-based model (R-EBM) is proposed to complement the auto-regressive ASR model to close the gap between the two distributions. Meanwhile, R-EBMs can also be regarded as utterance-level confidence estimators, which may benefit many downstream tasks. Experiments on a 100hr LibriSpeech dataset show that R-EBMs can reduce the word error rates (WERs) by 8.2%/6.7% while improving areas under precision-recall curves of confidence scores by 12.6%/28.4% on test-clean/test-other sets. Furthermore, on a state-of-the-art model using self-supervised learning (wav2vec 2.0), R-EBMs still significantly improves both the WER and confidence estimation performance.

Although end-to-end automatic speech recognition (e2e ASR) models are widely deployed in many applications, there have been very few studies to understand models' robustness against adversarial perturbations.In this paper, we explore whether a targeted universal perturbation vector exists for e2e ASR models.Our goal is to find perturbations that can mislead the models to predict the given targeted transcript such as "thank you" or empty string on any input utterance.We study two different attacks, namely additive and prepending perturbations, and their performances on the state-of-the-art LAS, CTC and RNN-T models.We find that LAS is the most vulnerable to perturbations among the three models.RNN-T is more robust against additive perturbations, especially on long utterances.And CTC is robust against both additive and prepending perturbations.To attack RNN-T, we find prepending perturbation is more effective than the additive perturbation, and can mislead the models to predict the same short target on utterances of arbitrary length.

Confidence scores are very useful for downstream applications of automatic speech recognition (ASR) systems. Recent works have proposed using neural networks to learn word or utterance confidence scores for end-to-end ASR. In those studies, word confidence by itself does not model deletions, and utterance confidence does not take advantage of word-level training signals. This paper proposes to jointly learn word confidence, word deletion, and utterance confidence. Empirical results show that multi-task learning with all three objectives improves confidence metrics (NCE, AUC, RMSE) without the need for increasing the model size of the confidence estimation module. Using the utterance-level confidence for rescoring also decreases the word error rates on Google's Voice Search and Long-tail Maps datasets by 3-5% relative, without needing a dedicated neural rescorer.

We present an approach for unsupervised learning of speech representation disentangling contents and styles. Our model consists of: (1) a local encoder that captures per-frame information; (2) a global encoder that captures per-utterance information; and (3) a conditional decoder that reconstructs speech given local and global latent variables. Our experiments show that (1) the local latent variables encode speech contents, as reconstructed speech can be recognized by ASR with low word error rates (WER), even with a different global encoding; (2) the global latent variables encode speaker style, as reconstructed speech shares speaker identity with the source utterance of the global encoding. Additionally, we demonstrate an useful application from our pre-trained model, where we can train a speaker recognition model from the global latent variables and achieve high accuracy by fine-tuning with as few data as one label per speaker.

In recent years, all-neural end-to-end approaches have obtained state-of-the-art results on several challenging automatic speech recognition (ASR) tasks. However, most existing works focus on building ASR models where train and test data are drawn from the same domain. This results in poor generalization characteristics on mismatched-domains: e.g., end-to-end models trained on short segments perform poorly when evaluated on longer utterances. In this work, we analyze the generalization properties of streaming and non-streaming recurrent neural network transducer (RNN-T) based end-to-end models in order to identify model components that negatively affect generalization performance. We propose two solutions: combining multiple regularization techniques during training, and using dynamic overlapping inference. On a long-form YouTube test set, when the non-streaming RNN-T model is trained with shorter segments of data, the proposed combination improves word error rate (WER) from 22.3% to 14.8%; when the streaming RNN-T model trained on short Search queries, the proposed techniques improve WER on the YouTube set from 67.0% to 25.3%. Finally, when trained on Librispeech, we find that dynamic overlapping inference improves WER on YouTube from 99.8% to 33.0%.

The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the over-all viability of the technology. In an effort to help align the research community’s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open.

This paper introduces a new end-to-end text-to-speech (E2E-TTS) toolkit named ESPnet-TTS, which is an extension of the open-source speech processing toolkit ESPnet. The toolkit supports state-of- the-art E2E-TTS models, including Tacotron 2, Transformer TTS, and FastSpeech, and also provides recipes inspired by the Kaldi automatic speech recognition (ASR) toolkit. The recipes are based on the design unified with the ESPnet ASR recipe, providing high reproducibility. The toolkit also provides pre-trained models and samples of all of the recipes so that users can use it as a baseline. Furthermore, the unified design enables the integration of ASR functions with TTS, e.g., ASR-based objective evaluation and semi- supervised learning with both ASR and TTS models. This paper describes the design of the toolkit and experimental evaluation in comparison with other toolkits. The experimental results show that our models can achieve state-of-the-art performance comparable to the other latest toolkits, resulting in a mean opinion score (MOS) of 4.25 on the LJSpeech dataset. The toolkit is publicly available at https://github.com/espnet/espnet.

In this paper, we propose to use pre-trained features from end-to-end ASR models to solve speech sentiment analysis as a down-stream task. We show that end-to-end ASR features, which integrate both acoustic and text information from speech, achieve promising results. We use RNN with self-attention as the sentiment classifier, which also provides an easy visualization through attention weights to help interpret model predictions. We use well benchmarked IEMOCAP dataset and a new large-scale speech sentiment dataset SWBD-sentiment for evaluation. Our approach improves the-state-of-the-art accuracy on IEMOCAP from 66.6% to 71.7%, and achieves an accuracy of 70.10% on SWBD-sentiment with more than 49,500 utterances.

Recently, SpecAugment, an augmentation scheme for automatic speech recognition that acts directly on the spectrogram of input utterances, has shown to be highly effective in enhancing the performance of end-to-end networks on public datasets. In this paper, we demonstrate its effectiveness on tasks with large scale datasets by investigating its application to the Google Multidomain Dataset (Narayanan et al., 2018). We achieve improvement across all test domains by mixing raw training data augmented with SpecAugment and noise-perturbed training data when training the acoustic model. We also introduce a modification of SpecAugment that adapts the time mask size and/or multiplicity depending on the length of the utterance, which can potentially benefit large scale tasks. By using adaptive masking, we are able to further improve the performance of the Listen, Attend and Spell model on LibriSpeech to 2.2% WER on test-clean and 5.2% WER on test-other.

Thus far, end-to-end (E2E) models have not been shown to outperform state-of-the-art conventional models with respect to both quality, i.e., word error rate (WER), and latency, i.e., the time the hypothesis is finalized after the user stops speaking. In this paper, we develop a first-pass Recurrent Neural Network Transducer (RNN-T) model and a second-pass Listen, Attend, Spell (LAS) rescorer that surpasses a conventional model in both quality and latency. On the quality side, we incorporate a large number of utterances across varied domains [1] to increase acoustic diversity and the vocabulary seen by the model. We also train with accented English speech to make the model more robust to different pronunciations. In addition, given the increased amount of training data, we explore a varied learning rate schedule. On the latency front, we explore using the end-of-sentence decision emitted by the RNN-T model to close the microphone, and also introduce various optimizations to improve the speed of LAS rescoring. Overall, we find that RNN-T+LAS offers a better WER and latency tradeoff compared to a conventional model. For example, for the same latency, RNN-T+LAS obtains a 8% relative improvement in WER, while being more than 400-times smaller in model size.

Recent neural text-to-speech (TTS) models with fine-grained latent features enable precise control of the prosody of synthesized speech. Such models typically incorporate a fine-grained variational autoencoder (VAE) structure, extracting latent features at each input token (e.g., phonemes). However, generating samples with the standard VAE prior often results in unnatural and discontinuous speech, with dramatic prosodic variation between tokens. This paper proposes a sequential prior in a discrete latent space which can generate more naturally sounding samples. This is accomplished by discretizing the latent features using vector quantization (VQ), and separately training an autoregressive (AR) prior model over the result. We evaluate the approach using listening tests, objective metrics of automatic speech recognition (ASR) performance, and measurements of prosody attributes. Experimental results show that the proposed model significantly improves the naturalness in random sample generation. Furthermore, initial experiments demonstrate that randomly sampling from the proposed model can be used as data augmentation to improve the ASR performance.

This paper proposes a hierarchical, fine-grained and interpretable latent variable model for prosody based on the Tacotron 2 text-to-speech model. It achieves multi-resolution modeling of prosody by conditioning finer level representations on coarser level ones. Additionally, it imposes hierarchical conditioning across all latent dimensions using a conditional variational auto-encoder (VAE) with an auto-regressive structure. Evaluation of reconstruction performance illustrates that the new structure does not degrade the model while allowing better interpretability. Interpretations of prosody attributes are provided together with the comparison between word-level and phone-level prosody representations. Moreover, both qualitative and quantitative evaluations are used to demonstrate the improvement in the disentanglement of the latent dimensions.

Speech synthesis has advanced to the point of being close to indistinguishable from human speech. However, efforts to train speech recognition systems on synthesized utterances have not been able to show that synthesized data can be effectively used to augment or replace human speech. In this work, we demonstrate that promoting consistent predictions in response to real and synthesized speech enables significantly improved speech recognition performance. We also find that training on 460 hours of LibriSpeech augmented with 500 hours of transcripts (without audio) performance is within 0.2% WER of a system trained on 960 hours of transcribed audio. This suggests that with this approach, when there is sufficient text available, reliance on transcribed audio can be cut nearly in half.

Recurrent Neural Network Transducer (RNN-T) models [1] for automatic speech recognition (ASR) provide high accuracy speech recognition. Such end-to-end (E2E) models combine acoustic, pronunciation and language models (AM, PM, LM) of a conventional ASR system into a single neural network, dramatically reducing complexity and model size.In this paper, we propose a technique for incorporating contextual signals, such as intelligent assistant device state or dialog state, directly into RNN-T models. We explore different encoding methods and demonstrate that RNN-T models can effectively utilize such context. Our technique results in reduction in Word Error Rate (WER) of up to 10.4% relative on a variety of contextual recognition tasks. We also demonstrate that proper regularization can be used to model context independently for improved overall quality.

Recently Transformer and Convolution neural network (CNN) based models have shown promising results in Automatic Speech Recognition (ASR), outperforming Recurrent neural networks (RNNs).Transformer models are good at capturing content-based global interactions, while CNNs exploit local features effectively.In this work, we achieve the best of both worlds by studying how to combine convolution neural networks and transformers to model both local and global dependencies of an audio sequence in a parameter-efficient way.To this regard, we propose the convolution-augmented transformer for speech recognition, named Conformer.Conformer significantly outperforms the previous Transformer and CNN based models achieving state-of-the-art accuracies.On the widely used LibriSpeech benchmark, our model achieves WER of 2.1%/4.3%without using a language model and 1.9%/3.9%with an external language model on test/testother.We also observe competitive performance of 2.7%/6.3%with a small model of only 10M parameters.

Convolutional neural networks (CNN) have shown promising results for end-to-end speech recognition, albeit still behind RNN/transformer based models in performance.In this paper, we study how to bridge this gap and go beyond with a novel CNN-RNN-transducer architecture, which we call ContextNet.ContextNet features a fully convolutional encoder that incorporates global context information into convolution layers by adding squeeze-and-excitation modules.In addition, we propose a simple scaling method that scales the widths of Con-textNet that achieves good trade-off between computation and accuracy.We demonstrate that on the widely used Librispeech benchmark, ContextNet achieves a word error rate (WER) of 2.1%/4.6%without external language model (LM), 1.9%/4.1% with LM and 2.9%/7.0%with only 10M parameters on the clean/noisy LibriSpeech test sets.This compares to the best previously published model of 2.0%/4.6%with LM and 3.9%/11.3%with 20M parameters.The superiority of the proposed ContextNet model is also verified on a much larger internal dataset.

Recently, a semi-supervised learning method known as "noisy student training" has been shown to improve image classification performance of deep networks significantly. Noisy student training is an iterative self-training method that leverages augmentation to improve network performance. In this work, we adapt and improve noisy student training for automatic speech recognition, employing (adaptive) SpecAugment as the augmentation method. We find effective methods to filter, balance and augment the data generated in between self-training iterations. By doing so, we are able to obtain word error rates (WERs) 4.2%/8.6% on the clean/noisy LibriSpeech test sets by only using the clean 100h subset of LibriSpeech as the supervised set and the rest (860h) as the unlabeled set. Furthermore, we are able to achieve WERs 1.7%/3.4% on the clean/noisy LibriSpeech test sets by using the unlab-60k subset of LibriLight as the unlabeled set for LibriSpeech 960h. We are thus able to improve upon the previous state-of-the-art clean/noisy test WERs achieved on LibriSpeech 100h (4.74%/12.20%) and LibriSpeech (1.9%/4.1%).

We present a multimodal corpus for sentiment analysis based on the existing Switchboard-1 Telephone Speech Corpus released by the Linguistic Data Consortium. This corpus extends the Switchboard-1 Telephone Speech Corpus by adding sentiment labels from 3 different human annotators for every transcript segment. Each sentiment label can be one of three options: positive, negative, and neutral. Annotators are recruited using Google Cloud’s data labeling service and the labeling task was conducted over the internet. The corpus contains a total of 49500 labeled speech segments covering 140 hours of audio. To the best of our knowledge, this is the largest multimodal Corpus for sentiment analysis that includes both speech and text features.

End-to-end automatic speech recognition (ASR) models, including both attention-based models and the recurrent neural network transducer (RNN-T), have shown superior performance compared to conventional systems [1], [2]. However, previous studies have focused primarily on short utterances that typically last for just a few seconds or, at most, a few tens of seconds. Whether such architectures are practical on long utterances that last from minutes to hours remains an open question. In this paper, we both investigate and improve the performance of end-to-end models on long-form transcription. We first present an empirical comparison of different end-to-end models on a real world long-form task and demonstrate that the RNN-T model is much more robust than attention-based systems in this regime. We next explore two improvements to attention-based systems that significantly improve its performance: restricting the attention to be monotonic, and applying a novel decoding algorithm that breaks long utterances into shorter overlapping segments. Combining these two improvements, we show that attention-based end-to-end models can be very competitive to RNN-T on long-form speech recognition.

Recent success of the Tacotron speech synthesis architecture and its variants in producing natural sounding multi-speaker synthesized speech has raised the exciting possibility of replacing expensive, manually transcribed, domain-specific, human speech that is used to train speech recognizers. The multi-speaker speech synthesis architecture can learn latent embedding spaces of prosody, speaker and style variations derived from input acoustic representations thereby allowing for manipulation of the synthesized speech. In this paper, we evaluate the feasibility of enhancing speech recognition performance using speech synthesis using two corpora from different domains. We explore algorithms to provide the necessary acoustic and lexical diversity needed for robust speech recognition. Finally, we demonstrate the feasibility of this approach as a data augmentation strategy for domain-transfer. We find that improvements to speech recognition performance is achievable by augmenting training data with synthesized material. However, there remains a substantial gap in performance between recognizers trained on human speech those trained on synthesized speech.

Although end-to-end text-to-speech (TTS) models such as Tacotron have shown excellent results, they typically require a sizable set of high-qualitypairs for training, which are expensive to collect. In this paper, we propose a semi-supervised training framework to improve the data efficiency of Tacotron. The idea is to allow Tacotron to utilize textual and acoustic knowledge contained in large, publicly-available text and speech corpora. Importantly, these external data are unpaired and potentially noisy. Specifically, first we embed each word in the input text into word vectors and condition the Tacotron encoder on them. We then use an unpaired speech corpus to pre-train the Tacotron decoder in the acoustic domain. Finally, we fine-tune the model using available paired data. We demonstrate that the proposed framework enables Tacotron to generate intelligible speech using less than half an hour of paired training data.

This paper presents a method to train end-to-end automatic speech recognition (ASR) models using unpaired data. Although the end-to-end approach can eliminate the need for expert knowledge such as pronunciation dictionaries to build ASR systems, it still requires a large amount of paired data, i.e., speech utterances and their transcriptions. Cycle-consistency losses have been recently proposed as a way to mitigate the problem of limited paired data. These approaches compose a reverse operation with a given transformation, e.g., text-to-speech (TTS) with ASR, to build a loss that only requires unsupervised data, speech in this example. Applying cycle consistency to ASR models is not trivial since fundamental information, such as speaker traits, are lost in the intermediate text bottleneck. To solve this problem, this work presents a loss that is based on the speech encoder state sequence instead of the raw speech signal. This is achieved by training a Text-To-Encoder model and defining a loss based on the encoder reconstruction error. Experimental results on the LibriSpeech corpus show that the proposed cycle-consistency training reduced the word error rate by 14.7% from an initial model trained with 100-hour paired data, using an additional 360 hours of audio data without transcriptions. We also investigate the use of text-only data mainly for language modeling to further improve the performance in the unpaired data training scenario.

To leverage crowd-sourced data to train multi-speaker text-to-speech (TTS) models that can synthesize clean speech for all speakers, it is essential to learn disentangled representations which can independently control the speaker identity and background noise in generated signals. However, learning such representations can be challenging, due to the lack of labels describing the recording conditions of each training example, and the fact that speakers and recording conditions are often correlated, e.g. since users often make many recordings using the same equipment. This paper proposes three components to address this problem by: (1) formulating a conditional generative model with factorized latent variables, (2) using data augmentation to add noise that is not correlated with speaker identity and whose label is known during training, and (3) using adversarial factorization to improve disentanglement. Experimental results demonstrate that the proposed method can disentangle speaker and noise attributes even if they are correlated in the training data, and can be used to consistently synthesize clean speech for all speakers. Ablation studies verify the importance of each proposed component.

We present two end-to-end models: Audio-to-Byte (A2B) and Byte-to-Audio (B2A), for multilingual speech recognition and synthesis. Prior work has predominantly used characters, sub-words or words as the unit of choice to model text. These units are difficult to scale to languages with large vocabularies, particularly in the case of multilingual processing. In this work, we model text via a sequence of Unicode bytes, specifically, the UTF-8 variable length byte sequence for each character. Bytes allow us to avoid large softmaxes in languages with large vocabularies, and share representations in multilingual models. We show that bytes are superior to grapheme characters over a wide variety of languages in monolingual end-to-end speech recognition. Additionally, our multilingual byte model outperform each respective single language baseline on average by 4.4% relatively. In Japanese-English code-switching speech, our multilingual byte model outperform our monolingual baseline by 38.6% relatively. Finally, we present an end-to-end multilingual speech synthesis model using byte representations which matches the performance of our monolingual baselines.

We present SpecAugment, a simple data augmentation method for speech recognition. SpecAugment is applied directly to the feature inputs of a neural network (i.e., filter bank coefficients). The augmentation policy consists of warping the features, masking blocks of frequency channels, and masking blocks of time steps. We apply SpecAugment on Listen, Attend and Spell networks for end-to-end speech recognition tasks. We achieve state-of-the-art performance on the LibriSpeech 960h and Swichboard 300h tasks, outperforming all prior work. On LibriSpeech, we achieve 6.8% WER on test-other without the use of a language model, and 5.8% WER with shallow fusion with a language model. This compares to the previous state-of-the-art hybrid system of 7.5% WER. For Switchboard, we achieve 7.2%/14.6% on the Switchboard/CallHome portion of the Hub5'00 test set without the use of a language model, and 6.8%/14.1% with shallow fusion, which compares to the previous state-of-the-art hybrid system at 8.3%/17.3% WER.

This paper introduces a new speech corpus called "LibriTTS" designed for text-to-speech use.It is derived from the original audio and text materials of the LibriSpeech corpus, which has been used for training and evaluating automatic speech recognition systems.The new corpus inherits desired properties of the LibriSpeech corpus while addressing a number of issues which make LibriSpeech less than ideal for text-to-speech work.The released corpus consists of 585 hours of speech data at 24kHz sampling rate from 2,456 speakers and the corresponding texts.Experimental results show that neural end-to-end TTS models trained from the LibriTTS corpus achieved above 4.0 in mean opinion scores in naturalness in five out of six evaluation speakers.

We present a multispeaker, multilingual text-to-speech (TTS) synthesis model based on Tacotron that is able to produce high quality speech in multiple languages.Moreover, the model is able to transfer voices across languages, e.g.synthesize fluent Spanish speech using an English speaker's voice, without training on any bilingual or parallel examples.Such transfer works across distantly related languages, e.g.English and Mandarin.Critical to achieving this result are: 1. using a phonemic input representation to encourage sharing of model capacity across languages, and 2. incorporating an adversarial loss term to encourage the model to disentangle its representation of speaker identity (which is perfectly correlated with language in the training data) from the speech content.Further scaling up the model by training on multiple speakers of each language, and incorporating an autoencoding input to help stabilize attention during training, results in a model which can be used to consistently synthesize intelligible speech for training speakers in all languages seen during training, and in native or foreign accents.

Common recurrent neural architectures scale poorly due to the intrinsic difficulty in parallelizing their state computations.In this work, we propose the Simple Recurrent Unit (SRU), a light recurrent unit that balances model capacity and scalability.SRU is designed to provide expressive recurrence, enable highly parallelized implementation, and comes with careful initialization to facilitate training of deep models.We demonstrate the effectiveness of SRU on multiple NLP tasks.SRU achieves 5-9x speed-up over cuDNN-optimized LSTM on classification and question answering datasets, and delivers stronger results than LSTM and convolutional models.We also obtain an average of 0.7 BLEU improvement over the Transformer model (Vaswani et al., 2017) on translation by incorporating SRU into the architecture.1

This paper describes Tacotron 2, a neural network architecture for speech synthesis directly from text. The system is composed of a recurrent sequence-to-sequence feature prediction network that maps character embeddings to mel-scale spectrograms, followed by a modified WaveNet model acting as a vocoder to synthesize time-domain waveforms from those spectrograms. Our model achieves a mean opinion score (MOS) of 4.53 comparable to a MOS of 4.58 for professionally recorded speech. To validate our design choices, we present ablation studies of key components of our system and evaluate the impact of using mel spectrograms as the conditioning input to WaveNet instead of linguistic, duration, and F0 features. We further show that using this compact acoustic intermediate representation allows for a significant reduction in the size of the WaveNet architecture.

In this work, we propose “global style tokens” (GSTs), a bank of embeddings that are jointly trained within Tacotron, a state-of-the-art end-to-end speech synthesis system. The embeddings are trained with no explicit labels, yet learn to model a large range of acoustic expressiveness. GSTs lead to a rich set of significant results. The soft interpretable “labels” they generate can be used to control synthesis in novel ways, such as varying speed and speaking style – independently of the text content. They can also be used for style transfer, replicating the speaking style of a single audio clip across an entire long-form text corpus. When trained on noisy, unlabeled found data, GSTs learn to factorize noise and speaker identity, providing a path towards highly scalable but robust speech synthesis.

We describe a neural network-based system for text-to-speech (TTS) synthesis that is able to generate speech audio in the voice of many different speakers, including those unseen during training. Our system consists of three independently trained components: (1) a speaker encoder network, trained on a speaker verification task using an independent dataset of noisy speech from thousands of speakers without transcripts, to generate a fixed-dimensional embedding vector from seconds of reference speech from a target speaker; (2) a sequence-to-sequence synthesis network based on Tacotron 2, which generates a mel spectrogram from text, conditioned on the speaker embedding; (3) an auto-regressive WaveNet-based vocoder that converts the mel spectrogram into a sequence of time domain waveform samples. We demonstrate that the proposed model is able to transfer the knowledge of speaker variability learned by the discriminatively-trained speaker encoder to the new task, and is able to synthesize natural speech from speakers that were not seen during training. We quantify the importance of training the speaker encoder on a large and diverse speaker set in order to obtain the best generalization performance. Finally, we show that randomly sampled speaker embeddings can be used to synthesize speech in the voice of novel speakers dissimilar from those used in training, indicating that the model has learned a high quality speaker representation.

In this paper we propose a novel data augmentation method for attention-based end-to-end automatic speech recognition (E2E-ASR), utilizing a large amount of text which is not paired with speech signals. Inspired by the back-translation technique proposed in the field of machine translation, we build a neural text-to-encoder model which predicts a sequence of hidden states extracted by a pre-trained E2E-ASR encoder from a sequence of characters. By using hidden states as a target instead of acoustic features, it is possible to achieve faster attention learning and reduce computational cost, thanks to sub-sampling in E2E-ASR encoder, also the use of the hidden states can avoid to model speaker dependencies unlike acoustic features. After training, the text-to-encoder model generates the hidden states from a large amount of unpaired text, then E2E-ASR decoder is retrained using the generated hidden states as additional training data. Experimental evaluation using LibriSpeech dataset demonstrates that our proposed method achieves improvement of ASR performance and reduces the number of unknown words without the need for paired data.

Domain mismatch between training and testing can lead to significant degradation in performance in many machine learning scenarios. Unfortunately, this is not a rare situation for automatic speech recognition deployments in real-world applications. Research on robust speech recognition can be regarded as trying to overcome this domain mismatch issue. In this paper, we address the unsupervised domain adaptation problem for robust speech recognition, where both source and target domain speech are available, but word transcripts are only available for the source domain speech. We present novel augmentation-based methods that transform speech in a way that does not change the transcripts. Specifically, we first train a variational autoencoder on both source and target domain data (without supervision) to learn a latent representation of speech. We then transform nuisance attributes of speech that are irrelevant to recognition by modifying the latent representations, in order to augment labeled training data with additional data whose distribution is more similar to the target domain. The proposed method is evaluated on the CHiME-4 dataset and reduces the absolute word error rate (WER) by as much as 35% compared to the non-adapted baseline.

We present a state-of-the-art end-to-end Automatic Speech Recognition (ASR) model. We learn to listen and write characters with a joint Connectionist Temporal Classification (CTC) and attention-based encoder-decoder network. The encoder is a deep Convolutional Neural Network (CNN) based on the VGG network. The CTC network sits on top of the encoder and is jointly trained with the attention-based decoder. During the beam search process, we combine the CTC predictions, the attention-based decoder predictions and a separately trained LSTM language model. We achieve a 5-10\% error reduction compared to prior systems on spontaneous Japanese and Chinese speech, and our end-to-end model beats out traditional hybrid ASR systems.

An ability to model a generative process and learn a latent representation for speech in an unsupervised fashion will be crucial to process vast quantities of unlabelled speech data. Recently, deep probabilistic generative models such as Variational Autoencoders (VAEs) have achieved tremendous success in modeling natural images. In this paper, we apply a convolutional VAE to model the generative process of natural speech. We derive latent space arithmetic operations to disentangle learned latent representations. We demonstrate the capability of our model to modify the phonetic content or the speaker identity for speech segments using the derived operations, without the need for parallel supervisory data.

We present a factorized hierarchical variational autoencoder, which learns disentangled and interpretable representations from sequential data without supervision. Specifically, we exploit the multi-scale nature of information in sequential data by formulating it explicitly within a factorized hierarchical graphical model that imposes sequence-dependent priors and sequence-independent priors to different sets of latent variables. The model is evaluated on two speech corpora to demonstrate, qualitatively, its ability to transform speakers or linguistic content by manipulating different sets of latent variables; and quantitatively, its ability to outperform an i-vector baseline for speaker verification and reduce the word error rate by as much as 35% in mismatched train/test scenarios for automatic speech recognition tasks.

In real-world data, e.g., from Web forums, text is often contaminated with redundant or irrelevant content, which leads to introducing noise in machine learning algorithms. In this paper, we apply Long Short-Term Memory networks with an attention mechanism, which can select important parts of text for the task of similar question retrieval from community Question Answering (cQA) forums. In particular, we use the attention weights for both selecting entire sentences and their subparts, i.e., word/chunk, from shallow syntactic trees. More interestingly, we apply tree kernels to the filtered text representations, thus exploiting the implicit features of the subtree space for learning question reranking. Our results show that the attention-based pruning allows for achieving the top position in the cQA challenge of SemEval 2016, with a relatively large gap from the other participants while greatly decreasing running time.

Deep Neural Networks (DNNs) trained on multilingual data have proven useful for improving speech recognition in languages with limited resources. In this framework, data from rich resource languages are pooled together to train a single system and then adapted to a new language. However, data from a rich language that are similar to the target language are generally more helpful. We explore methods of training bottleneck features by using data that are more similar to the target language. Our experiments on speech recognition and keyword spotting tasks with IARPA-Babel languages show that our proposed methods outperform typical multilingual DNNs.

Although great progress has been made in automatic speech recognition (ASR), significant performance degradation still exists in distant talking scenarios due to significantly lower signal power. In this paper, a novel adaptation framework, named integrated adaptation with multi-factor joint-learning, is proposed to improve the recognition accuracy for distant speech recognition. We explore and extract speaker, phone and environment factor representations using deep neural networks (DNNs), which are integrated into the main ASR DNN to improve classification accuracy. In addition, the hidden activations in the main ASR DNN are used to improve the factor extraction, which in turn helps the ASR DNN. All the model parameters, including those in the ASR DNN and factor extractor DNNs, are jointly optimized under the multi-task learning framework. Further more, unlike prior techniques, our novel approach requires no explicit separate stages for factor extraction and adaptation. Experiments on the AMI single distant microphone (SDM) task show that the proposed architecture can significantly reduce word error rate (WER) and additional improvement can be achieved by combining it with the i-vector adaptation. Our best configuration obtained more than 15% and 10% relative reduction on WER over the baselines using the SDM and close-talk data generated alignments, respectively.

Long Short-Term Memory (LSTM) is a particular type of recurrent neural network (RNN) that can model long term temporal dynamics. Recently it has been shown that LSTM-RNNs can achieve higher recognition accuracy than deep feed-forword neural networks (DNNs) in acoustic modelling. However, speaker adaption for LSTM-RNN based acoustic models has not been well investigated. In this paper, we study the LSTM-RNN speaker-aware training that incorporates the speaker information during model training to normalise the speaker variability. We first present several speaker-aware training architectures, and then empirically evaluate three types of speaker representation: I-vectors, bottleneck speaker vectors and speaking rate. Furthermore, to factorize the variability in the acoustic signals caused by speakers and phonemes respectively, we investigate the speaker-aware and phone-aware joint training under the framework of multi-task learning. In AMI meeting speech transcription task, speaker-aware training of LSTM-RNNs reduces word error rates by 6.5% relative to a very strong LSTM-RNN baseline, which uses FMLLR features.

Despite the significant progress in speech recognition enabled by deep neural networks, poor performance persists in some scenarios. In this work, we focus on far-field speech recognition which remains challenging due to high levels of noise and reverberation in the captured speech signals. We propose to represent the stages of acoustic processing including beamforming, feature extraction, and acoustic modeling, as three components of a single unified computational network. The parameters of a frequency-domain beam-former are first estimated by a network based on features derived from the microphone channels. These filter coefficients are then applied to the array signals to form an enhanced signal. Conventional features are then extracted from this signal and passed to a second network that performs acoustic modeling for classification. The parameters of both the beamforming and acoustic modeling networks are trained jointly using back-propagation with a common cross-entropy objective function. In experiments on the AMI meeting corpus, we observed improvements by pre-training each sub-network with a network-specific objective function before joint training of both networks. The proposed method obtained a 3.2% absolute word error rate reduction compared to a conventional pipeline of independent processing stages.

In this paper, we investigate the use of prediction-adaptation-correction recurrent neural networks (PAC-RNNs) for low-resource speech recognition. A PAC-RNN is comprised of a pair of neural networks in which a correction network uses auxiliary information given by a prediction network to help estimate the state probability. The information from the correction network is also used by the prediction network in a recurrent loop. Our model outperforms other state-of-the-art neural networks (DNNs, LSTMs) on IARPA-Babel tasks. Moreover, transfer learning from a language that is similar to the target language can help improve performance further.

In this paper, we extend the deep long short-term memory (DL-STM) recurrent neural networks by introducing gated direct connections between memory cells in adjacent layers. These direct links, called highway connections, enable unimpeded information flow across different layers and thus alleviate the gradient vanishing problem when building deeper LSTMs. We further introduce the latency-controlled bidirectional LSTMs (BLSTMs) which can exploit the whole history while keeping the latency under control. Efficient algorithms are proposed to train these novel networks using both frame and sequence discriminative criteria. Experiments on the AMI distant speech recognition (DSR) task indicate that we can train deeper LSTMs and achieve better improvement from sequence training with highway LSTMs (HLSTMs). Our novel model obtains 43.9/47.7% WER on AMI (SDM) dev and eval sets, outperforming all previous works. It beats the strong DNN and DLSTM baselines with 15.7% and 5.3% relative improvement respectively.

Recurrent neural networks (RNNs) are naturally suitable for speech recognition because of their ability of utilizing dynamically changing temporal information. Deep RNNs have been argued to be able to model temporal relationships at different time granularities, but suffer vanishing gradient problems. In this paper, we extend stacked long short-term memory (LSTM) RNNs by using grid LSTM blocks that formulate computation along not only the temporal dimension, but also the depth dimension, in order to alleviate this issue. Moreover, we prioritize the depth dimension over the temporal one to provide the depth dimension more updated information, since the output from it will be used for classification. We call this model the prioritized Grid LSTM (pGLSTM). Extensive experiments on four large datasets (AMI, HKUST, GALE, and MGB) indicate that the pGLSTM outperforms alternative deep LSTM models, beating stacked LSTMs with 4% to 7% relative improvement, and achieve new benchmarks among uni-directional models on all datasets.

We propose the prediction-adaptation-correction RNN (PAC-RNN), in which a correction DNN estimates the state posterior probability based on both the current frame and the prediction made on the past frames by a prediction DNN. The result from the main DNN is fed back to the prediction DNN to make better predictions for the future frames. In the PAC-RNN, we can consider that, given the new, current frame information, the main DNN makes a correction on the prediction made by the prediction DNN. Alternatively, it can be viewed as adapting the main DNN's behavior based on the prediction DNN's prediction. Experiments on the TIMIT phone recognition task indicate that the PAC-RNN outperforms DNN, RNN, and LSTM with 2.4%, 2.1%, and 1.9% absolute phone accuracy improvement, respectively. We found that incorporating the prediction objective and including the recurrent loop are both important to boost the performance of the PAC-RNN.

Deep Neural Networks (DNN) have been largely used and successfully applied in the context of speaker independent Automatic Speech Recognition (ASR). However, these models are not easily adapted to model a specific speaker characteristic. Recently, one approach was proposed to address this issue, which consists of using the I-vector representation as input to the DNN. The I-vector is playing the role of providing information about the speaker as well as the environmental conditions for a given recording. This approach achieved a significant improvement in the context of a hybrid system of DNN combined with Hidden Markov Model (HMM). In this paper, we study the effect of speaker adaptation based on the I-vector framework in the context of stacked bottleneck features. These features, extracted from a second level of DNNs, are modelled by a classical Gaussian Mixture Model (GMM) ASR system. The proposed approach achieved an absolute WER improvement of 1.2% on an Arabic Broadcast news task. Index Terms: DNN, I-Vector, Bottleneck Features, Speech Recognition

In this paper, we investigate the use of deep neural networks (DNNs) to generate a stacked bottleneck (SBN) feature representation for low-resource speech recognition. We examine different SBN extraction architectures, and incorporate low-rank matrix factorization in the final weight layer. Experiments on several low-resource languages demonstrate the effectiveness of the SBN configurations when compared to state-of-the-art hybrid DNN approaches.

This paper describes a detailed comparison of several state-of-the-art speech recognition techniques applied to a limited Ara-bic broadcast news dataset. The different approaches were all trained on 50 hours of transcribed audio from the Al-Jazeera news channel. The best results were obtained using i-vector-based speaker adaptation in a training scenario using the Min-imum Phone Error (MPE) criteria combined with sequential Deep Neural Network (DNN) training. We report results for two different types of test data: broadcast news reports, with a best word error rate (WER) of 17.86%, and a broadcast conver-sations with a best WER of 29.85%. The overall WER on this test set is 25.6%. Index Terms: Arabic, ASR system, Kaldi 1.

In this paper, we explore multilingual feature-level data sharing via Deep Neural Network (DNN) stacked bottleneck features. Given a set of available source languages, we apply language identification to pick the language most similar to the target language, for more efficient use of multilingual resources. Our experiments with IARPA-Babel languages show that bottleneck features trained on the most similar source language perform better than those trained on all available source languages. Further analysis suggests that only data similar to the target language is useful for multilingual training. Index Terms: Multilingual, Bottleneck features, DNN

Acoustic feature similarity between search results has been shown to be very helpful for the task of spoken term detection (STD). A graph-based re-ranking approach for STD has been proposed based on the concept that search results, which are acoustically similar to other results with higher confidence scores, should have higher scores themselves. In this approach, the similarity between all search results of a given term are considered as a graph, and the confidence scores of the search results propagate through this graph. Since this approach can improve STD results without any additional labelled data, it is especially suitable for STD on languages with limited amounts of annotated data. However, its performance has not been widely studied on benchmark corpora. In this paper, we investigate the effectiveness of the graph-based reranking approach on limited language data from the IARPA Babel program. Experiments on the low-resource languages, Assamese, Bengali and Lao, show that graph-based re-ranking improves STD systems using fuzzy matching, and lattices based on different kinds of units including words, subwords, and hybrids. Index Terms: Random Walk, Spoken Term Detection

Neural network based approaches have recently produced record-setting performances in natural language understanding tasks such as word labeling. In the word labeling task, a tagger is used to assign a label to each word in an input sequence. Specifically, simple recurrent neural networks (RNNs) and convolutional neural networks (CNNs) have shown to significantly outperform the previous state-of-the-art - conditional random fields (CRFs). This paper investigates using long short-term memory (LSTM) neural networks, which contain input, output and forgetting gates and are more advanced than simple RNN, for the word labeling task. To explicitly model output-label dependence, we propose a regression model on top of the LSTM un-normalized scores. We also propose to apply deep LSTM to the task. We investigated the relative importance of each gate in the LSTM by setting other gates to a constant and only learning particular gates. Experiments on the ATIS dataset validated the effectiveness of the proposed models.

The creation of a pronunciation lexicon remains the most inefficient process in developing an Automatic Speech Recognizer (ASR).In this paper, we propose an unsupervised alternative -requiring no language-specific knowledge -to the conventional manual approach for creating pronunciation dictionaries.We present a hierarchical Bayesian model, which jointly discovers the phonetic inventory and the Letter-to-Sound (L2S) mapping rules in a language using only transcribed data.When tested on a corpus of spontaneous queries, the results demonstrate the superiority of the proposed joint learning scheme over its sequential counterpart, in which the latent phonetic inventory and L2S mappings are learned separately.Furthermore, the recognizers built with the automatically induced lexicon consistently outperform grapheme-based recognizers and even approach the performance of recognition systems trained using conventional supervised procedures.