ASR - VLSP 2021: Automatic Speech Recognition with Blank Label Re-weighting

The Recurrent Neural Network Transducer (RNN-T) extends Connectionist Temporal Classification (CTC) by jointly modeling both input-output and output-output dependencies, and it has been successfully applied in end-to-end speech recognition. RNN-T adds language information by expanding its dimensions in this way, but the model training is difficult. And some paths of the model are unreasonable when decoding. Therefore, we present the Alignment RNN Transducer (ART) algorithm, which uses the forward-backward algorithm in CTC to find the best path alignment information as context-related information for training. We not only reduce the training dimensions but also adds context information to the model. And we still use CTC as the loss function to avoid decoding the impossible path in the RNN-T. We verify the algorithm on a Mandarin speech corpus AIShell-1, and it achieves a 13.65% CER, compared with the 16.79% CER given by the RNN-T model.

The recurrent neural network transducer (RNN-T) has recently become the mainstream end-to-end approach for streaming automatic speech recognition (ASR). To estimate the output distributions over subword units, RNN-T uses a fully connected layer as the joint network to fuse the acoustic representations extracted using the acoustic encoder with the text representations obtained using the prediction network based on the previous subword units. In this paper, we propose to use gating, bilinear pooling, and a combination of them in the joint network to produce more expressive representations to feed into the output layer. A regularisation method is also proposed to enable better acoustic encoder training by reducing the gradients back-propagated into the prediction network at the beginning of RNN-T training. Experimental results on a multilingual ASR setting for voice search over nine languages show that the joint use of the proposed methods can result in 4%--5% relative word error rate reductions with only a few million extra parameters.

This article describes a density ratio approach to integrating external Language Models (LMs) into end-to-end models for Automatic Speech Recognition (ASR). Applied to a Recurrent Neural Network Transducer (RNN-T) ASR model trained on a given domain, a matched in-domain RNN-LM, and a target domain RNN-LM, the proposed method uses Bayes' Rule to define RNN-T posteriors for the target domain, in a manner directly analogous to the classic hybrid model for ASR based on Deep Neural Networks (DNNs) or LSTMs in the Hidden Markov Model (HMM) framework (Bourlard & Morgan, 1994). The proposed approach is evaluated in cross-domain and limited-data scenarios, for which a significant amount of target domain text data is used for LM training, but only limited (or no) {audio, transcript} training data pairs are used to train the RNN-T. Specifically, an RNN-T model trained on paired audio & transcript data from YouTube is evaluated for its ability to generalize to Voice Search data. The Density Ratio method was found to consistently outperform the dominant approach to LM and end-to-end ASR integration, Shallow Fusion.

End-to-end (E2E) models fold the acoustic, pronunciation and language models of a conventional speech recognition model into one neural network with a much smaller number of parameters than a conventional ASR system, thus making it suitable for on-device applications. For example, recurrent neural network transducer (RNN-T) as a streaming E2E model has shown promising potential for on-device ASR [1]. For such applications, quality and latency are two critical factors. We propose to reduce E2E model's latency by extending the RNN-T endpointer (RNN-T EP) model [2] with additional early and late penalties. By further applying the minimum word error rate (MWER) training technique [3], we achieved 8.0% relative word error rate (WER) reduction and 130ms 90-percentile latency reduction over [2] on a Voice Search test set. We also experimented with a second-pass Listen, Attend and Spell (LAS) rescorer [4]. Although it did not directly improve the first pass latency, the large WER reduction provides extra room to trade WER for latency. RNN-T EP+LAS, together with MWER training brings in 18.7% relative WER reduction and 160ms 90-percentile latency reductions compared to the original proposed RNN-T EP [2] model.

An investigation of neural uncertainty estimation for target speaker extraction equipped RNN transducer

In this paper we present an end-to-end speech recognition system that can recognize single-channel speech where multiple talkers can speak at the same time (overlapping speech) by using a neural network model based on Recurrent Neural Network Transducer (RNN-T) architecture. We augment the conventional RNN-T architecture by including a masking model for separation of encoded audio features, and multiple label encoders to encode transcripts from different speakers. We use a masking L2 loss to prevent transcripts to align to wrong speakers’ audio, and a speaker embedding loss to facilitate speaker tracking. We show that by using these additional training objectives, the proposed augmented RNN-T model can be trained with simulated overlapping speech data and can achieve a WER of 32% on words in overlapping speech segments from real-life telephone conversations. Our analysis of manual transcription task on the same test set shows that transcribing overlapping speech is hard even for humans who can get a WER of 37% compared to ground-truth.

The recurrent neural network transducer (RNN-T) model has been proved effective for keyword spotting (KWS) recently. However, compared with cross-entropy (CE) or connectionist temporal classification (CTC) based models, the additional prediction network in the RNN-T model increases the model size and computational cost. Besides, since the keyword training data usually only contain the keyword sequence, the prediction network might has over-fitting problems. In this paper, we improve the RNN-T modeling for small-footprint keyword spotting in three aspects. First, to address the over-fitting issue, we explore multi-task training where a CTC loss is added to the encoder. The CTC loss is calculated with both KWS data and ASR data, while the RNN-T loss is calculated with ASR data so that only the encoder is augmented with KWS data. Second, we use the feed-forward neural network to replace the LSTM for prediction network modeling. Thus all possible prediction network outputs could be pre-computed for decoding. Third, we further improve the model with transfer learning, where a model trained with 160 thousand hours of ASR data is used to initialize the KWS model. On a self-collected far-field wake-word testset, the proposed RNN-T system greatly improves the performance comparing with a strong "keyword-filler" baseline.

In multi-talker scenarios such as meetings and conversations, speech processing systems are usually required to transcribe the audio as well as identify the speakers for downstream applications. Since overlapped speech is common in this case, conventional approaches usually address this problem in a cascaded fashion that involves speech separation, speech recognition and speaker identification that are trained independently. In this paper, we propose Streaming Unmixing, Recognition and Identification Transducer (SURIT) -- a new framework that deals with this problem in an end-to-end streaming fashion. SURIT employs the recurrent neural network transducer (RNN-T) as the backbone for both speech recognition and speaker identification. We validate our idea on the LibrispeechMix dataset -- a multi-talker dataset derived from Librispeech, and present encouraging results.

Recurrent neural networks (RNNs) can be used to handle sequential patterns and have been used for speech recognition. To overcome the shortcomings of RNN, recurrent sub neural networks (RSNNs) are used, where an RSNN is built independently for each class. The training algorithm of the RSNN is based on the backpropagation algorithm. Speaker dependent connected Chinese digit-speech recognition experiments were carried out. Some factors influencing the performance of RSNNs have been studied. The experiments show that RSNN is easier to train and gives higher performance than RNN.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

We introduce two techniques, length perturbation and n-best based label smoothing, to improve generalization of deep neural network (DNN) acoustic models for automatic speech recognition (ASR). Length perturbation is a data augmentation algorithm that randomly drops and inserts frames of an utterance to alter the length of the speech feature sequence. N-best based label smoothing randomly injects noise to ground truth labels during training in order to avoid overfitting, where the noisy labels are generated from n-best hypotheses. We evaluate these two techniques extensively on the 300-hour Switchboard (SWB300) dataset and an in-house 500-hour Japanese (JPN500) dataset using recurrent neural network transducer (RNNT) acoustic models for ASR. We show that both techniques improve the generalization of RNNT models individually and they can also be complementary. In particular, they yield good improvements over a strong SWB300 baseline and give state-of-art performance on SWB300 using RNNT models.

We present Bifocal RNN-T, a new variant of the Recurrent Neural Network Transducer (RNN-T) architecture designed for improved inference time latency on speech recognition tasks. The architecture enables a dynamic pivot for its runtime compute pathway, namely taking advantage of keyword spotting to select which component of the network to execute for a given audio frame. To accomplish this, we leverage a recurrent cell we call the Bifocal LSTM (BF-LSTM), which we detail in the paper. The architecture is compatible with other optimization strategies such as quantization, sparsification, and applying time-reduction layers, making it especially applicable for deployed, real-time speech recognition settings. We present the architecture and report comparative experimental results on voice-assistant speech recognition tasks. Specifically, we show our proposed Bifocal RNN-T can improve inference cost by 29.1% with matching word error rates and only a minor increase in memory size.

In this paper, we present a comparative study on the robustness of two different online streaming speech recognition models: Monotonic Chunkwise Attention (MoChA) and Recurrent Neural Network-Transducer (RNN-T). We explore three recently proposed data augmentation techniques, namely, multi-conditioned training using an acoustic simulator, Vocal Tract Length Perturbation (VTLP) for speaker variability, and SpecAugment. Experimental results show that unidirectional models are in general more sensitive to noisy examples in the training set. It is observed that the final performance of the model depends on the proportion of training examples processed by data augmentation techniques. MoChA models generally perform better than RNN-T models. However, we observe that training of MoChA models seems to be more sensitive to various factors such as the characteristics of training sets and the incorporation of additional augmentations techniques. On the other hand, RNN-T models perform better than MoChA models in terms of latency, inference time, and the stability of training. Additionally, RNN-T models are generally more robust against noise and reverberation. All these advantages make RNN-T models a better choice for streaming on-device speech recognition compared to MoChA models.

In this paper, a novel approach for implementing Arabic isolated speech recognition is described. While most of the literature on speech recognition (SR) is based on hidden Markov models (HMM), the present system is implemented by modular recurrent Elman neural networks (MRENN). The promising results obtained through this design show that this new neural networks approach can compete with the traditional HMM-based speech recognition approaches.

Deep learning methodologies have had a major impact on performance across a wide variety of machine learning tasks, and speech recognition is no exception. We describe a set of deep learning techniques that proved to be particularly successful in achieving performance gains in word error rate on a popular large vocabulary conversational speech recognition benchmark task (“Switchboard”). We found that the best performance is achieved by combining features from both recurrent and convolutional neural networks. We compare two recurrent architectures: partially unfolded nets with max-out activations and bidirectional long short-term memory nets. In addition, inspired by the success of convolutional networks for image classification, we designed a convolutional net with many convolutional layers and small kernels that create a receptive field with more nonlinearity and fewer parameters than standard configurations. When combined, these neural networks achieve a word error rate of 6.2% on this difficult task; this was the best reported rate at the time of this writing and is even more remarkable given that human performance itself is estimated to be 4% on this data.

In this paper, the improvements in the recently implemented Kannada speech recognition system is demonstrated in detail. The Kannada automatic speech recognition (ASR) system consists of ASR models which are created by using Kaldi, IVRS call flow and weather and agricultural commodity prices information databases. The task specific speech data used in the recently developed spoken dialogue system had high level of different background noises. The different types of noises present in collected speech data had an adverse effect on the on line and off line speech recognition performances. Therefore, to improve the speech recognition accuracy in Kannada ASR system, a noise reduction algorithm is developed which is a fusion of spectral subtraction with voice activity detection (SS-VAD) and minimum mean square error spectrum power estimator based on zero crossing (MMSE-SPZC) estimator. The noise elimination algorithm is added in the system before the feature extraction part. An alternative ASR models are created using subspace Gaussian mixture models (SGMM) and deep neural network (DNN) modeling techniques. The experimental results show that, the fusion of noise elimination technique and SGMM/DNN based modeling gives a better relative improvement of 7.68% accuracy compared to the recently developed GMM-HMM based ASR system. The least word error rate (WER) acoustic models could be used in spoken dialogue system. The developed spoken query system is tested from Karnataka farmers under uncontrolled environment.