We introduce a framework for analyzing various types of information in an NLP Transformer. In this approach, we distinguish four layers of information: positional, syntactic, semantic, and contextual. We also argue that the common practice of adding positional information to semantic embedding is sub-optimal and propose instead a Linear-and-Add approach. Our analysis reveals an autogenetic separation of positional information through the deep layers. We show that the distilled positional components of the embedding vectors follow the path of a helix, both on the encoder side and on the decoder side. We additionally show that on the encoder side, the conceptual dimensions generate Part-of-Speech (PoS) clusters. On the decoder side, we show that a di-gram approach helps to reveal the PoS clusters of the next token. Our approach paves a way to elucidate the processing of information through the deep layers of an NLP Transformer.
GPU decoding significantly accelerates the output of ASR predictions. While GPUs are already being used for online ASR decoding, post-processing and rescoring on GPUs have not been properly investigated yet. Rescoring with available contextual information can considerably improve ASR predictions. Previous studies have proven the viability of lattice rescoring in decoding and biasing language model (LM) weights in offline and online CPU scenarios. In real-time GPU decoding, partial recognition hypotheses are produced without lattice generation, which makes the implementation of biasing more complex. The paper proposes and describes an approach to integrate contextual biasing in real-time GPU decoding while exploiting the standard Kaldi GPU decoder. Besides the biasing of partial ASR predictions, our approach also permits dynamic context switching allowing a flexible rescoring per each speech segment directly on GPU. The code is publicly released and tested with open-sourced test sets.
In this paper, we describe VivesDebate-Speech, a corpus of spoken argumentation created to leverage audio features for argument mining tasks. The creation of this corpus represents an important contribution to the intersection of speech processing and argument mining communities, and one of the most complete publicly available resources in this topic. Moreover, we have performed a set of first-of-their-kind experiments which show an improvement when integrating audio features into the argument mining pipeline. The provided results can be used as a baseline for future research.
This paper introduces SpeeChain, an open-source Pytorch-based toolkit designed to develop the machine speech chain for large-scale use. This first release focuses on the TTS-to-ASR chain, a core component of the machine speech chain, that refers to the TTS data augmentation by unspoken text for ASR. To build an efficient pipeline for the large-scale TTS-to-ASR chain, we implement easy-to-use multi-GPU batch-level model inference, multi-dataloader batch generation, and on-the-fly data selection techniques. In this paper, we first explain the overall procedure of the TTS-to-ASR chain and the difficulties of each step. Then, we present a detailed ablation study on different types of unlabeled data, data filtering thresholds, batch composition, and real-synthetic data ratios. Our experimental results on train_clean_460 of LibriSpeech demonstrate that our TTS-to-ASR chain can significantly improve WER in a semi-supervised setting.
Inspired by humans comprehending speech in a multi-modal manner, various audio-visual datasets have been constructed. However, most existing datasets focus on English, induce dependencies with various prediction models during dataset preparation, and have only a small number of multi-view videos. To mitigate the limitations, we recently developed the Open Large-scale Korean Audio-Visual Speech (OLKAVS) dataset, which is the largest among publicly available audio-visual speech datasets. The dataset contains 1,150 hours of transcribed audio from 1,107 Korean speakers in a studio setup with nine different viewpoints and various noise situations. We also provide the pre-trained baseline models for two tasks, audio-visual speech recognition and lip reading. We conducted experiments based on the models to verify the effectiveness of multi-modal and multi-view training over uni-modal and frontal-view-only training. We expect the OLKAVS dataset to facilitate multi-modal research in broader areas such as Korean speech recognition, speaker recognition, pronunciation level classification, and mouth motion analysis.
End-to-end automatic speech recognition (E2E-ASR) has the potential to improve performance, but a specific issue that needs to be addressed is the difficulty it has in handling enharmonic words: named entities (NEs) with the same pronunciation and part of speech that are spelled differently. This often occurs with Japanese personal names that have the same pronunciation but different Kanji characters. Since such NE words tend to be important keywords, ASR easily loses user trust if it misrecognizes them. To solve these problems, this paper proposes a novel retraining-free customized method for E2E-ASRs based on a named-entity-aware E2E-ASR model and phoneme similarity estimation. Experimental results show that the proposed method improves the target NE character error rate by 35.7% on average relative to the conventional E2E-ASR model when selecting personal names as a target NE.
Recently, many efforts have been made to explore how the brain processes speech using electroencephalographic (EEG) signals, where deep learning-based approaches were shown to be applicable in this field. In order to decode speech signals from EEG signals, linear networks, convolutional neural networks (CNN) and long short-term memory networks are often used in a supervised manner. Recording EEG-speech labeled data is rather time-consuming and laborious, while unlabeled EEG data is abundantly available. Whether self-supervised methods are helpful to learn EEG representation to boost the performance of EEG auditory-related tasks has not been well explored. In this work, we first propose a self-supervised model based on contrastive loss and reconstruction loss to learn EEG representations, and then use the obtained pre-trained model as a feature extractor for downstream tasks. Second, for two considered downstream tasks, we use CNNs and Transformer networks to learn local features and global features, respectively. Finally, the EEG data from other channels are mixed into the chosen EEG data for augmentation. The effectiveness of our method is verified on the EEG match-mismatch and EEG regression tasks of the ICASSP2023 Auditory EEG Challenge.
Auditory attention decoding (AAD) is a technique used to identify and amplify the talker that a listener is focused on in a noisy environment. This is done by comparing the listener's brainwaves to a representation of all the sound sources to find the closest match. The representation is typically the waveform or spectrogram of the sounds. The effectiveness of these representations for AAD is uncertain. In this study, we examined the use of self-supervised learned speech representation in improving the accuracy and speed of AAD. We recorded the brain activity of three subjects using invasive electrocorticography (ECoG) as they listened to two conversations and focused on one. We used WavLM to extract a latent representation of each talker and trained a spatiotemporal filter to map brain activity to intermediate representations of speech. During the evaluation, the reconstructed representation is compared to each speaker's representation to determine the target speaker. Our results indicate that speech representation from WavLM provides better decoding accuracy and speed than the speech envelope and spectrogram. Our findings demonstrate the advantages of self-supervised learned speech representation for auditory attention decoding and pave the way for developing brain-controlled hearable technologies.
Self-supervised speech pre-training enables deep neural network models to capture meaningful and disentangled factors from raw waveform signals. The learned universal speech representations can then be used across numerous downstream tasks. These representations, however, are sensitive to distribution shifts caused by environmental factors, such as noise and/or room reverberation. Their large sizes, in turn, make them unfeasible for edge applications. In this work, we propose a knowledge distillation methodology termed RobustDistiller which compresses universal representations while making them more robust against environmental artifacts via a multi-task learning objective. The proposed layer-wise distillation recipe is evaluated on top of three well-established universal representations, as well as with three downstream tasks. Experimental results show the proposed methodology applied on top of the WavLM Base+ teacher model outperforming all other benchmarks across noise types and levels, as well as reverberation times. Oftentimes, the obtained results with the student model (24M parameters) achieved results inline with those of the teacher model (95M).
The interest in employing automatic speech recognition (ASR) in applications for reading practice has been growing in recent years. In a previous study, we presented an ASR-based Dutch reading tutor application that was developed to provide instantaneous feedback to first-graders learning to read. We saw that ASR has potential at this stage of the reading process, as the results suggested that pupils made progress in reading accuracy and fluency by using the software. In the current study, we used children's speech from an existing corpus (JASMIN) to develop two new ASR systems, and compared the results to those of the previous study. We analyze correct/incorrect classification of the ASR systems using human transcripts at word level, by means of evaluation measures such as Cohen's Kappa, Matthews Correlation Coefficient (MCC), precision, recall and F-measures. We observe improvements for the newly developed ASR systems regarding the agreement with human-based judgment and correct rejection (CR). The accuracy of the ASR systems varies for different reading tasks and word types. Our results suggest that, in the current configuration, it is difficult to classify isolated words. We discuss these results, possible ways to improve our systems and avenues for future research.