Self-supervised learning (SSL) speech models such as wav2vec and HuBERT have demonstrated state-of-the-art performance on automatic speech recognition (ASR) and proved to be extremely useful in low label-resource settings. However, the success of SSL models has yet to transfer to utterance-level tasks such as speaker, emotion, and language recognition, which still require supervised fine-tuning of the SSL models to obtain good performance. We argue that the problem is caused by the lack of disentangled representations and an utterance-level learning objective for these tasks. Inspired by how HuBERT uses clustering to discover hidden acoustic units, we formulate a factor analysis (FA) model that uses the discovered hidden acoustic units to align the SSL features. The underlying utterance-level representations are disentangled from the content of speech using probabilistic inference on the aligned features. Furthermore, the variational lower bound derived from the FA model provides an utterance-level objective, allowing error gradients to be backpropagated to the Transformer layers to learn highly discriminative acoustic units. When used in conjunction with HuBERT's masked prediction training, our models outperform the current best model, WavLM, on all utterance-level non-semantic tasks on the SUPERB benchmark with only 20% of labeled data.
MIT Computer Science and Artificial Intelligence Laboratory（麻省理工学院计算机科学与人工智能实验室）；Le Mans University（勒芒大学）
We propose the SAMU-XLSR: Semantically-Aligned Multimodal Utterance-level Cross-Lingual Speech Representation learning framework. Unlike previous works on speech representation learning, which learns multilingual contextual speech embedding at the resolution of an acoustic frame (10-20ms), this work focuses on learning multimodal (speech-text) multilingual speech embedding at the resolution of a sentence (5-10s) such that the embedding vector space is semantically aligned across different languages. We combine state-of-the-art multilingual acoustic frame-level speech representation learning mo…
Stuttering is a speech disorder where the natural flow of speech is interrupted by blocks, repetitions or prolongations of syllables, words and phrases. The majority of existing automatic speech recognition (ASR) interfaces perform poorly on utterances with stutter, mainly due to lack of matched training data. Synthesis of speech with stutter thus presents an opportunity to improve ASR for this type of speech. We describe Stutter-TTS, an end-to-end neural text-to-speech model capable of synthesizing diverse types of stuttering utterances. We develop a simple, yet effective prosody-control stra…
National Taiwan Normal University（国立台湾师范大学）；Chunghwa Telecom Laboratories（中华电信研究所）
Conversational speech normally is embodied with loose syntactic structures at the utterance level but simultaneously exhibits topical coherence relations across consecutive utterances. Prior work has shown that capturing longer context information with a recurrent neural network or long short-term memory language model (LM) may suffer from the recent bias while excluding the long-range context. In order to capture the long-term semantic interactions among words and across utterances, we put forward disparate conversation history fusion methods for language modeling in automatic speech recognit…
We discuss the problem of echographic transcription in autoregressive sequence-to-sequence attentional architectures for automatic speech recognition, where a model produces very long sequences of repetitive outputs when presented with out-of-domain utterances. We decode audio from the British National Corpus with an attentional encoder-decoder model trained solely on the LibriSpeech corpus. We observe that there are many 5-second recordings that produce more than 500 characters of decoding output (i.e. more than 100 characters per second). A frame-synchronous hybrid (DNN-HMM) model trained on…
Recent studies of streaming automatic speech recognition (ASR) recurrent neural network transducer (RNN-T)-based systems have fed the encoder with past contextual information in order to improve its word error rate (WER) performance. In this paper, we first propose a contextual-utterance training technique which makes use of the previous and future contextual utterances in order to do an implicit adaptation to the speaker, topic and acoustic environment. Also, we propose a dual-mode contextual-utterance training technique for streaming automatic speech recognition (ASR) systems. This proposed …
Aalborg University（奥尔堡大学）；Tencent AI Lab（腾讯人工智能实验室）；Microsoft Research（微软研究院）；Oticon A/S（奥迪康公司）
In this paper we propose the utterance-level Permutation Invariant Training (uPIT) technique. uPIT is a practically applicable, end-to-end, deep learning based solution for speaker independent multi-talker speech separation. Specifically, uPIT extends the recently proposed Permutation Invariant Training (PIT) technique with an utterance-level cost function, hence eliminating the need for solving an additional permutation problem during inference, which is otherwise required by frame-level PIT. We achieve this using Recurrent Neural Networks (RNNs) that, during training, minimize the utterance-…
A novel framework for meeting transcription using asynchronous microphones is proposed in this paper. It consists of audio synchronization, speaker diarization, utterance-wise speech enhancement using guided source separation, automatic speech recognition, and duplication reduction. Doing speaker diarization before speech enhancement enables the system to deal with overlapped speech without considering sampling frequency mismatch between microphones. Evaluation on our real meeting datasets showed that our framework achieved a character error rate (CER) of 28.7 % by using 11 distributed microph…
Although deep learning-based end-to-end Automatic Speech Recognition (ASR) has shown remarkable performance in recent years, it suffers severe performance regression on test samples drawn from different data distributions. Test-time Adaptation (TTA), previously explored in the computer vision area, aims to adapt the model trained on source domains to yield better predictions for test samples, often out-of-domain, without accessing the source data. Here, we propose the Single-Utterance Test-time Adaptation (SUTA) framework for ASR, which is the first TTA study on ASR to our best knowledge. The …
We introduce a new approach for speech pre-training named SPIRAL which works by learning denoising representation of perturbed data in a teacher-student framework. Specifically, given a speech utterance, we first feed the utterance to a teacher network to obtain corresponding representation. Then the same utterance is perturbed and fed to a student network. The student network is trained to output representation resembling that of the teacher. At the same time, the teacher network is updated as moving average of student's weights over training steps. In order to prevent representation collapse…
In order to build language technologies for majority of the languages, it is important to leverage the resources available in public domain on the internet - commonly referred to as `Found Data'. However, such data is characterized by the presence of non-standard, non-trivial variations. For instance, speech resources found on the internet have non-speech content, such as music. Therefore, speech recognition and speech synthesis models need to be robust to such variations. In this work, we present an analysis to show that it is important to disentangle the latent causal factors of variation in…
We present a framework for building speech-to-text translation (ST) systems using only monolingual speech and text corpora, in other words, speech utterances from a source language and independent text from a target language. As opposed to traditional cascaded systems and end-to-end architectures, our system does not require any labeled data (i.e., transcribed source audio or parallel source and target text corpora) during training, making it especially applicable to language pairs with very few or even zero bilingual resources. The framework initializes the ST system with a cross-modal biling…
How to effectively incorporate cross-utterance information cues into a neural language model (LM) has emerged as one of the intriguing issues for automatic speech recognition (ASR). Existing research efforts on improving contextualization of an LM typically regard previous utterances as a sequence of additional input and may fail to capture complex global structural dependencies among these utterances. In view of this, we in this paper seek to represent the historical context information of an utterance as graph-structured data so as to distill cross-utterances, global word interaction relatio…
Utterance-Level Methods for Identifying Reliable ASR-Output for Child Speech
Centre for Language Studies, Radboud University, Netherlands（语言研究所以及鲁文大学，荷兰）；Department of Language and Communication, Radboud University, Netherlands（语言与交流系，鲁文大学，荷兰）
Confidence measure is a performance index of particular importance for automatic speech recognition (ASR) systems deployed in real-world scenarios. In the present study, utterance-level neural confidence measure (NCM) in end-to-end automatic speech recognition (E2E ASR) is investigated. The E2E system adopts the joint CTC-attention Transformer architecture. The prediction of NCM is formulated as a task of binary classification, i.e., accept/reject the input utterance, based on a set of predictor features acquired during the ASR decoding process. The investigation is focused on evaluating and c…
In recent studies, self-supervised pre-trained models tend to outperform supervised pre-trained models in transfer learning. In particular, self-supervised learning (SSL) of utterance-level speech representation can be used in speech applications that require discriminative representation of consistent attributes within an utterance: speaker, language, emotion, and age. Existing frame-level self-supervised speech representation, e.g., wav2vec, can be used as utterance-level representation with pooling, but the models are usually large. There are also SSL techniques to learn utterance-level rep…
Transformer-based models have demonstrated their effectiveness in automatic speech recognition (ASR) tasks and even shown superior performance over the conventional hybrid framework. The main idea of Transformers is to capture the long-range global context within an utterance by self-attention layers. However, for scenarios like conversational speech, such utterance-level modeling will neglect contextual dependencies that span across utterances. In this paper, we propose to explicitly model the inter-sentential information in a Transformer based end-to-end architecture for conversational speec…
National Taiwan University（台湾大学）；Academia Sinica（中央研究院）
Speech enhancement model is used to map a noisy speech to a clean speech. In the training stage, an objective function is often adopted to optimize the model parameters. However, in most studies, there is an inconsistency between the model optimization criterion and the evaluation criterion on the enhanced speech. For example, in measuring speech intelligibility, most of the evaluation metric is based on a short-time objective intelligibility (STOI) measure, while the frame based minimum mean square error (MMSE) between estimated and clean speech is widely used in optimizing the model. Due to …
Alibaba Group（阿里巴巴集团）；Nanyang Technological University（南洋理工大学）；Alibaba-NTU Singapore Joint Research Institute（阿里巴巴-新加坡南洋理工大学联合研究院）
Existing self-supervised pre-trained speech models have offered an effective way to leverage massive unannotated corpora to build good automatic speech recognition (ASR). However, many current models are trained on a clean corpus from a single source, which tends to do poorly when noise is present during testing. Nonetheless, it is crucial to overcome the adverse influence of noise for real-world applications. In this work, we propose a novel training framework, called deHuBERT, for noise reduction encoding inspired by H. Barlow's redundancy-reduction principle. The new framework improves the …
This paper introduces a new open-source speech corpus named "speechocean762" designed for pronunciation assessment use, consisting of 5000 English utterances from 250 non-native speakers, where half of the speakers are children. Five experts annotated each of the utterances at sentence-level, word-level and phoneme-level. A baseline system is released in open source to illustrate the phoneme-level pronunciation assessment workflow on this corpus. This corpus is allowed to be used freely for commercial and non-commercial purposes. It is available for free download from OpenSLR, and the correspo…
In this thesis, we present a statistical language model for resolving speech repairs, intonational boundaries and discourse markers. Rather than finding the best word interpretation for an acoustic signal, we redefine the speech recognition problem to so that it also identifies the POS tags, discourse markers, speech repairs and intonational phrase endings (a major cue in determining utterance units). Adding these extra elements to the speech recognition problem actually allows it to better predict the words involved, since we are able to make use of the predictions of boundary tones, discours…
Considering the abundance of unlabeled speech data and the high labeling costs, unsupervised learning methods can be essential for better system development. One of the most successful methods is contrastive self-supervised methods, which require negative sampling: sampling alternative samples to contrast with the current sample (anchor). However, it is hard to ensure if all the negative samples belong to classes different from the anchor class without labels. This paper applies a non-contrastive self-supervised learning method on an unlabeled speech corpus to learn utterance-level embeddings.…
Indian Institute of Science (IISc)（印度科学学院）；IIIT Hyderabad（印度海得拉巴国际信息技术学院）
In this study, listeners of varied Indian nativities are asked to listen and recognize TIMIT utterances spoken by American speakers. We have three kinds of responses from each listener while they recognize an utterance: 1. Sentence difficulty ratings, 2. Speaker difficulty ratings, and 3. Transcription of the utterance. From these transcriptions, word error rate (WER) is calculated and used as a metric to evaluate the similarity between the recognized and the original sentences.The sentences selected in this study are categorized into three groups: Easy, Medium and Hard, based on the frequency…
University Grenoble Alpes（格勒诺布尔阿尔卑斯大学）；CNRS（法国国家科学研究中心）；Grenoble INP（格勒诺布尔理工学院）；LIG（格勒诺布尔信息实验室）
The main aim of this paper is to investigate automatic quality assessment for spoken language translation (SLT). More precisely, we investigate SLT errors that can be due to transcription (ASR) or to translation (MT) modules. This paper investigates automatic detection of SLT errors using a single classifier based on joint ASR and MT features. We evaluate both 2-class (good/bad) and 3-class (good/badASR/badMT ) labeling tasks. The 3-class problem necessitates to disentangle ASR and MT errors in the speech translation output and we propose two label extraction methods for this non trivial step.…
End-to-end models have achieved state-of-the-art results on several automatic speech recognition tasks. However, they perform poorly when evaluated on long-form data, e.g., minutes long conversational telephony audio. One reason the model fails on long-form speech is that it has only seen short utterances during training. In this paper we study the effect of training utterance length on the word error rate (WER) for RNN-transducer (RNN-T) model. We compare two widely used training objectives, log loss (or RNN-T loss) and minimum word error rate (MWER) loss. We conduct experiments on telephony …
This paper describes "TLT-school" a corpus of speech utterances collected in schools of northern Italy for assessing the performance of students learning both English and German. The corpus was recorded in the years 2017 and 2018 from students aged between nine and sixteen years, attending primary, middle and high school. All utterances have been scored, in terms of some predefined proficiency indicators, by human experts. In addition, most of utterances recorded in 2017 have been manually transcribed carefully. Guidelines and procedures used for manual transcriptions of utterances will be des…
This paper describes a novel knowledge distillation framework that leverages acoustically qualified speech data included in an existing training data pool as privileged information. In our proposed framework, a student network is trained with multiple soft targets for each utterance that consist of main soft targets from original speakers' utterance and alternative targets from other speakers' utterances spoken under better acoustic conditions as a secondary view. These qualified utterances from other speakers, used to generate better soft targets, are collected from a qualified data pool by u…
The University of Tokyo（东京大学）；Tokyo Institute of Technology（东京工业大学）
This paper presents sampling-based speech parameter generation using moment-matching networks for Deep Neural Network (DNN)-based speech synthesis. Although people never produce exactly the same speech even if we try to express the same linguistic and para-linguistic information, typical statistical speech synthesis produces completely the same speech, i.e., there is no inter-utterance variation in synthetic speech. To give synthetic speech natural inter-utterance variation, this paper builds DNN acoustic models that make it possible to randomly sample speech parameters. The DNNs are trained s…
The bootstrap resampling method has been popular for performing significance analysis on word error rate (WER) in automatic speech recognition (ASR) evaluation. To deal with dependent speech data, the blockwise bootstrap approach is also introduced. By dividing utterances into uncorrelated blocks, this approach resamples these blocks instead of original data. However, it is typically nontrivial to uncover the dependent structure among utterances and identify the blocks, which might lead to subjective conclusions in statistical testing. In this paper, we present graphical lasso based methods to…
Self-Supervised Learning (SSL) has made great strides recently. SSL speech models achieve decent performance on a wide range of downstream tasks, suggesting that they extract different aspects of information from speech. However, how SSL models store various information in hidden representations without interfering is still poorly understood. Taking the recently successful SSL model, HuBERT, as an example, we explore how the SSL model processes and stores speaker information in the representation. We found that HuBERT stores speaker information in representations whose positions correspond to …
Nara Institute of Science and Technology（奈良先端科学技术大学院大学）；Google LLC（谷歌有限责任公司）
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 styl…
Masked speech modeling (MSM) methods such as wav2vec2 or w2v-BERT learn representations over speech frames which are randomly masked within an utterance. While these methods improve performance of Automatic Speech Recognition (ASR) systems, they have one major limitation. They treat all unsupervised speech samples with equal weight, which hinders learning as not all samples have relevant information to learn meaningful representations. In this work, we address this limitation. We propose ask2mask (ATM), a novel approach to focus on specific samples during MSM pre-training. ATM employs an exter…
Dysarthria is a motor speech impairment affecting millions of people. Dysarthric speech can be far less intelligible than those of non-dysarthric speakers, causing significant communication difficulties. The goal of our work is to develop a model for dysarthric to healthy speech conversion using Cycle-consistent GAN. Using 18,700 dysarthric and 8,610 healthy control Korean utterances that were recorded for the purpose of automatic recognition of voice keyboard in a previous study, the generator is trained to transform dysarthric to healthy speech in the spectral domain, which is then converted…
Although modern automatic speech recognition (ASR) systems can achieve high performance, they may produce errors that weaken readers' experience and do harm to downstream tasks. To improve the accuracy and reliability of ASR hypotheses, we propose a cross-modal post-processing system for speech recognizers, which 1) fuses acoustic features and textual features from different modalities, 2) joints a confidence estimator and an error corrector in multi-task learning fashion and 3) unifies error correction and utterance rejection modules. Compared with single-modal or single-task models, our prop…
Language modeling (LM) for automatic speech recognition (ASR) does not usually incorporate utterance level contextual information. For some domains like voice assistants, however, additional context, such as the time at which an utterance was spoken, provides a rich input signal. We introduce an attention mechanism for training neural speech recognition language models on both text and non-linguistic contextual data. When applied to a large de-identified dataset of utterances collected by a popular voice assistant platform, our method reduces perplexity by 7.0% relative over a standard LM that…
We propose a novel Patched Multi-Condition Training (pMCT) method for robust Automatic Speech Recognition (ASR). pMCT employs Multi-condition Audio Modification and Patching (MAMP) via mixing {\it patches} of the same utterance extracted from clean and distorted speech. Training using patch-modified signals improves robustness of models in noisy reverberant scenarios. Our proposed pMCT is evaluated on the LibriSpeech dataset showing improvement over using vanilla Multi-Condition Training (MCT). For analyses on robust ASR, we employed pMCT on the VOiCES dataset which is a noisy reverberant data…
Mitsubishi Electric Research Laboratories (MERL)（三菱电机研究实验室）
This paper addresses end-to-end automatic speech recognition (ASR) for long audio recordings such as lecture and conversational speeches. Most end-to-end ASR models are designed to recognize independent utterances, but contextual information (e.g., speaker or topic) over multiple utterances is known to be useful for ASR. In our prior work, we proposed a context-expanded Transformer that accepts multiple consecutive utterances at the same time and predicts an output sequence for the last utterance, achieving 5-15% relative error reduction from utterance-based baselines in lecture and conversati…
NTT Media Intelligence Laboratories, NTT Corporation（日本电信电话株式会社媒体智能研究所）
We present a novel personalized voice activity detection (PVAD) learning method that does not require enrollment data during training. PVAD is a task to detect the speech segments of a specific target speaker at the frame level using enrollment speech of the target speaker. Since PVAD must learn speakers' speech variations to clarify the boundary between speakers, studies on PVAD used large-scale datasets that contain many utterances for each speaker. However, the datasets to train a PVAD model are often limited because substantial cost is needed to prepare such a dataset. In addition, we cann…
For conversational large-vocabulary continuous speech recognition (LVCSR) tasks, up to about two thousand hours of audio is commonly used to train state of the art models. Collection of labeled conversational audio however, is prohibitively expensive, laborious and error-prone. Furthermore, academic corpora like Fisher English (2004) or Switchboard (1992) are inadequate to train models with sufficient accuracy in the unbounded space of conversational speech. These corpora are also timeworn due to dated acoustic telephony features and the rapid advancement of colloquial vocabulary and idiomatic…
University College London（伦敦大学学院）；SoftV（SoftV公司）；Goldsmiths College, University of London（伦敦大学金史密斯学院）
Anomia (word-finding difficulties) is the hallmark of aphasia, an acquired language disorder most commonly caused by stroke. Assessment of speech performance using picture naming tasks is a key method for both diagnosis and monitoring of responses to treatment interventions by people with aphasia (PWA). Currently, this assessment is conducted manually by speech and language therapists (SLT). Surprisingly, despite advancements in automatic speech recognition (ASR) and artificial intelligence with technologies like deep learning, research on developing automated systems for this task has been sc…
In most automatic speech recognition (ASR) systems, the audio signal is processed to produce a time series of sensor measurements (e.g., filterbank outputs). This time series encodes semantic information in a speaker-dependent way. An earlier paper showed how to use the sequence of sensor measurements to derive an "inner" time series that is unaffected by any previous invertible transformation of the sensor measurements. The current paper considers two or more speakers, who mimic one another in the following sense: when they say the same words, they produce sensor states that are invertibly ma…
University of Wrocław（弗罗茨瓦夫大学）；Jacobs University Bremen（不来梅雅各布大学）；Université de Montréal（蒙特利尔大学）
Recurrent sequence generators conditioned on input data through an attention mechanism have recently shown very good performance on a range of tasks in- cluding machine translation, handwriting synthesis and image caption gen- eration. We extend the attention-mechanism with features needed for speech recognition. We show that while an adaptation of the model used for machine translation in reaches a competitive 18.7% phoneme error rate (PER) on the TIMIT phoneme recognition task, it can only be applied to utterances which are roughly as long as the ones it was trained on. We offer a qualitativ…
