This paper describes the recent development of ESPnet (https://github.com/espnet/espnet), an end-to-end speech processing toolkit. This project was initiated in December 2017 to mainly deal with end-to-end speech recognition experiments based on sequence-to-sequence modeling. The project has grown rapidly and now covers a wide range of speech processing applications. Now ESPnet also includes text to speech (TTS), voice conversation (VC), speech translation (ST), and speech enhancement (SE) with support for beamforming, speech separation, denoising, and dereverberation. All applications are trained in an end-to-end manner, thanks to the generic sequence to sequence modeling properties, and they can be further integrated and jointly optimized. Also, ESPnet provides reproducible all-in-one recipes for these applications with state-of-the-art performance in various benchmarks by incorporating transformer, advanced data augmentation, and conformer. This project aims to provide up-to-date speech processing experience to the community so that researchers in academia and various industry scales can develop their technologies collaboratively.

The 2020 ESPnet Update: New Features, Broadened Applications, Performance Improvements, and Future Plans

Connectionist Temporal Classification (CTC) is a widely used approach for automatic speech recognition (ASR) that performs conditionally independent monotonic alignment. However for translation, CTC exhibits clear limitations due to the contextual and non-monotonic nature of the task and thus lags behind attentional decoder approaches in terms of translation quality. In this work, we argue that CTC does in fact make sense for translation if applied in a joint CTC/attention framework wherein CTC’s core properties can counteract several key weaknesses of pure-attention models during training and decoding. To validate this conjecture, we modify the Hybrid CTC/Attention model originally proposed for ASR to support text-to-text translation (MT) and speech-to-text translation (ST). Our proposed joint CTC/attention models outperform pure-attention baselines across six benchmark translation tasks.

CTC Alignments Improve Autoregressive Translation

Recent end-to-end Automatic Speech Recognition (ASR) systems demonstrated the ability to outperform conventional hybrid DNN/ HMM ASR. Aside from architectural improvements in those systems, those models grew in terms of depth, parameters and model capacity. However, these models also require more training data to achieve comparable performance. 
In this work, we combine freely available corpora for German speech recognition, including yet unlabeled speech data, to a big dataset of over $1700$h of speech data. For data preparation, we propose a two-stage approach that uses an ASR model pre-trained with Connectionist Temporal Classification (CTC) to boot-strap more training data from unsegmented or unlabeled training data. Utterances are then extracted from label probabilities obtained from the network trained with CTC to determine segment alignments. With this training data, we trained a hybrid CTC/attention Transformer model that achieves $12.8\%$ WER on the Tuda-DE test set, surpassing the previous baseline of $14.4\%$ of conventional hybrid DNN/HMM ASR.

CTC-Segmentation of Large Corpora for German End-to-end Speech Recognition

Audio Adversarial Examples (AAE) represent purposefully designed inputs meant to trick Automatic Speech Recognition (ASR) systems into misclassification. The present work proposes MP3 compression as a means to decrease the impact of Adversarial Noise (AN) in audio samples transcribed by ASR systems. To this end, we generated AAEs with a new variant of the Fast Gradient Sign Method for an end-to-end, hybrid CTC-attention ASR system. The MP3’s effectiveness against AN is then validated by two objective indicators: (1) Character Error Rates (CER) that measure the speech decoding performance of four ASR models trained on different audio formats (both uncompressed and MP3-compressed) and (2) Signal-to-Noise Ratio (SNR) estimated for uncompressed and MP3-compressed AAEs that are reconstructed in the time domain by feature inversion. We found that MP3 compression applied to AAEs indeed reduces the CER when compared to uncompressed AAEs. Moreover, feature-inverted (reconstructed) AAEs had significantly higher SNRs after MP3 compression, indicating that AN was reduced. In contrast to AN, MP3 compression applied to utterances augmented with regular noise resulted in more transcription errors, giving further evidence that MP3 encoding is effective in diminishing AN exclusively.

MP3 Compression to Diminish Adversarial Noise in End-to-End Speech Recognition

Artificial Neural Networks (ANNs) were created inspired by the neural networks in the human brain and have been widely applied in speech processing. The application areas of ANN include: Speech recognition, speech emotion recognition, language identification, speech enhancement, and speech separation, amongst others. Likewise, given that speech processing performed by humans involves complex cognitive processes known as auditory attention, there has been a growing amount of papers proposing ANNs supported by deep learning algorithms in conjunction with some mechanism to achieve symmetry with the human attention process. However, while these ANN approaches include attention, there is no categorization of attention integrated into the deep learning algorithms and their relation with human auditory attention. Therefore, we consider it necessary to have a review of the different ANN approaches inspired in attention to show both academic and industry experts the available models for a wide variety of applications. Based on the PRISMA methodology, we present a systematic review of the literature published since 2000, in which deep learning algorithms are applied to diverse problems related to speech processing. In this paper 133 research works are selected and the following aspects are described: (i) Most relevant features, (ii) ways in which attention has been implemented, (iii) their hypothetical relationship with human attention, and (iv) the evaluation metrics used. Additionally, the four publications most related with human attention were analyzed and their strengths and weaknesses were determined.

Attention-Inspired Artificial Neural Networks for Speech Processing: A Systematic Review

Recent end-to-end Automatic Speech Recognition (ASR) systems demonstrated the ability to outperform conventional hybrid DNN/HMM ASR. Aside from architectural improvements in those systems, those models grew in terms of depth, parameters and model capacity. However, these models also require more training data to achieve comparable performance.

CTC-Segmentation of Large Corpora for German End-to-End Speech Recognition

Feature distillation always leads to significant performance improvements, but requires extra training budgets. To address the problem, we propose TFD, a simple and effective Teacher-Free Distillation framework, which seeks to reuse the privileged features within the student network itself. Specifically, TFD squeezes feature knowledge in the deeper layers into the shallow ones by minimizing feature loss. Thanks to the narrow gap of these self-features, TFD only needs to adopt a simple l2 loss without complex transformations. Extensive experiments on recognition benchmarks show that our framework can achieve superior performance than teacher-based feature distillation methods. On the ImageNet dataset, our approach achieves 0.8% gains for ResNet18, which surpasses other state-of-the-art training techniques.

Teacher-free Distillation via Regularizing Intermediate Representation

Knowledge distillation conditioned on intermediate feature representations always leads to significant performance improvements. Conventional feature distillation framework demands extra selecting/training budgets of teachers and complex transformations to align the features between teacher-student models. To address the problem, we analyze teacher roles in feature distillation and have an intriguing observation: additional teacher architectures are not always necessary. Then we propose Tf-FD, a simple yet effective Teacher- $$\boldsymbol{f}$$ ree Feature Distillation framework, reusing channel-wise and layer-wise meaningful features within the student to provide teacher-like knowledge without an additional model. In particular, our framework is subdivided into intra-layer and inter-layer distillation. The intra-layer Tf-FD performs feature salience ranking and transfers the knowledge from salient feature to redundant feature within the same layer. For inter-layer Tf-FD, we deal with distilling high-level semantic knowledge embedded in the deeper layer representations to guide the training of shallow layers. Benefiting from the small gap between these self-features, Tf-FD simply needs to optimize extra feature mimicking losses without complex transformations. Furthermore, we provide insightful discussions to shed light on Tf-FD from feature regularization perspectives. Our experiments conducted on classification and object detection tasks demonstrate that our technique achieves state-of-the-art results on different models with fast training speeds. Code is available at https://lilujunai.github.io/Teacher-free-Distillation/ . 

Self-Regulated Feature Learning via Teacher-free Feature Distillation

Existing feature distillation methods commonly adopt the One-to-one Representation Matching between any pre-selected teacher-student layer pair. In this paper, we present N-to-One Representation (NORM), a new two-stage knowledge distillation method, which relies on a simple Feature Transform (FT) module consisting of two linear layers. In view of preserving the intact information learnt by the teacher network, during training, our FT module is merely inserted after the last convolutional layer of the student network. The first linear layer projects the student representation to a feature space having N times feature channels than the teacher representation from the last convolutional layer, and the second linear layer contracts the expanded output back to the original feature space. By sequentially splitting the expanded student representation into N non-overlapping feature segments having the same number of feature channels as the teacher's, they can be readily forced to approximate the intact teacher representation simultaneously, formulating a novel many-to-one representation matching mechanism conditioned on a single teacher-student layer pair. After training, such an FT module will be naturally merged into the subsequent fully connected layer thanks to its linear property, introducing no extra parameters or architectural modifications to the student network at inference. Extensive experiments on different visual recognition benchmarks demonstrate the leading performance of our method. For instance, the ResNet18|MobileNet|ResNet50-1/4 model trained by NORM reaches 72.14%|74.26%|68.03% top-1 accuracy on the ImageNet dataset when using a pre-trained ResNet34|ResNet50|ResNet50 model as the teacher, achieving an absolute improvement of 2.01%|4.63%|3.03% against the individually trained counterpart. Code is available at https://github.com/OSVAI/NORM

Lujun Li

Papers

CTC-Segmentation of Large Corpora for German End-to-End Speech Recognition

CTC-Segmentation of Large Corpora for German End-to-end Speech Recognition

Teacher-free Distillation via Regularizing Intermediate Representation

Self-Regulated Feature Learning via Teacher-free Feature Distillation

NORM: Knowledge Distillation via N-to-One Representation Matching