This work addresses the persistent challenges of substantial training time and GPU resource requirements even when training lightweight encoder-vocoder models for Textless NLP. We reduce training steps significantly while improving performance by a) leveraging learning rate schedulers for efficient and faster convergence b) optimizing hop length and c) tuning the interpolation scale factors for better audio quality. Additionally, we explore the latent space representation for Indian languages such as Tamil and Bengali for the acoustic unit discovery and voice conversion task. Our approach leverages a quantized encoder architecture, in conjunction with a vocoder which utilizes the proposed mixture of optimized hop length, tuned interpolation scale factors and a cyclic learning rate scheduler. We obtain consistently good results across English, Tamil and Bengali datasets. The proposed method excels in capturing complex linguistic patterns, resulting in clear reconstructed audio during voice conversion with significantly reduced training time.
Self-supervised speech representation learning has become essential for extracting meaningful features from untranscribed audio. Recent advances highlight the potential of deriving discrete symbols from the features correlated with linguistic units, which enables text-less training across diverse tasks. In particular, sentence-level Self-Distillation of the pretrained HuBERT (SD-HuBERT) induces syllabic structures within latent speech frame representations extracted from an intermediate Transformer layer. In SD-HuBERT, sentence-level representation is accumulated from speech frame features through self-attention layers using a special CLS token. However, we observe that the information aggregated in the CLS token correlates more with speaker identity than with linguistic content. To address this, we propose a speech-only self-supervised fine-tuning approach that separates syllabic units from speaker information. Our method introduces speaker perturbation as data augmentation and adopts a frame-level training objective to prevent the CLS token from aggregating paralinguistic information. Experimental results show that our approach surpasses the current state-of-the-art method in most syllable segmentation and syllabic unit quality metrics on Librispeech, underscoring its effectiveness in promoting syllabic organization within speech-only models.
End-to-end (E2E) keyword search (KWS) has emerged as an alternative and complimentary approach to conventional keyword search which depends on the output of automatic speech recognition (ASR) systems. While E2E methods greatly simplify the KWS pipeline, they generally have worse performance than their ASR-based counterparts, which can benefit from pretraining with untranscribed data. In this work, we propose a method for pretraining E2E KWS systems with untranscribed data, which involves using acoustic unit discovery (AUD) to obtain discrete units for untranscribed data and then learning to locate sequences of such units in the speech. We conduct experiments across languages and AUD systems: we show that finetuning such a model significantly outperforms a model trained from scratch, and the performance improvements are generally correlated with the quality of the AUD system used for pretraining.
We revisit a self-supervised method that segments unlabelled speech into word-like segments. We start from the two-stage duration-penalised dynamic programming method that performs zero-resource segmentation without learning an explicit lexicon. In the first acoustic unit discovery stage, we replace contrastive predictive coding features with HuBERT. After word segmentation in the second stage, we get an acoustic word embedding for each segment by averaging HuBERT features. These embeddings are clustered using K-means to get a lexicon. The result is good full-coverage segmentation with a lexicon that achieves state-of-the-art performance on the ZeroSpeech benchmarks.
Toward high-performance multilingual automatic speech recognition (ASR), various types of linguistic information and model design have demonstrated their effectiveness independently. They include language identity (LID), phoneme information, language-specific processing modules, and cross-lingual self-supervised speech representation. It is expected that leveraging their benefits synergistically in a unified solution would further improve the overall system performance. This paper presents a novel design of a hierarchical information path, named LUPET, which sequentially encodes, from the shallow layers to deep layers, multiple aspects of linguistic and acoustic information at diverse granularity scales. The path starts from LID prediction, followed by acoustic unit discovery, phoneme sharing, and finally token recognition routed by a mixture-of-expert. ASR experiments are carried out on 10 languages in the Common Voice corpus. The results demonstrate the superior performance of LUPET as compared to the baseline systems. Most importantly, LUPET effectively mitigates the issue of performance compromise of high-resource languages with low-resource ones in the multilingual setting.
Self-supervised speech representation models have succeeded in various tasks, but improving them for content-related problems using unlabeled data is challenging. We propose speaker-invariant clustering (Spin), a novel self-supervised learning method that clusters speech representations and performs swapped prediction between the original and speaker-perturbed utterances. Spin disentangles speaker information and preserves content representations with just 45 minutes of fine-tuning on a single GPU. Spin improves pre-trained networks and outperforms prior methods in speech recognition and acoustic unit discovery.
Self-supervised methods such as Contrastive predictive Coding (CPC) have greatly improved the quality of the unsupervised representations. These representations significantly reduce the amount of labeled data needed for downstream task performance, such as automatic speech recognition. CPC learns representations by learning to predict future frames given current frames. Based on the observation that the acoustic information, e.g., phones, changes slower than the feature extraction rate in CPC, we propose regularization techniques that impose slowness constraints on the features. Here we propose two regularization techniques: Self-expressing constraint and Left-or-Right regularization. We evaluate the proposed model on ABX and linear phone classification tasks, acoustic unit discovery, and automatic speech recognition. The regularized CPC trained on 100 hours of unlabeled data matches the performance of the baseline CPC trained on 360 hours of unlabeled data. We also show that our regularization techniques are complementary to data augmentation and can further boost the system's performance. In monolingual, cross-lingual, or multilingual settings, with/without data augmentation, regardless of the amount of data used for training, our regularized models outperformed the baseline CPC models on the ABX task.
Recent progress in self-supervised or unsupervised machine learning has opened the possibility of building a full speech processing system from raw audio without using any textual representations or expert labels such as phonemes, dictionaries or parse trees. The contribution of the Zero Resource Speech Challenge series since 2015 has been to break down this long-term objective into four well-defined tasks -- Acoustic Unit Discovery, Spoken Term Discovery, Discrete Resynthesis, and Spoken Language Modeling -- and introduce associated metrics and benchmarks enabling model comparison and cumulative progress. We present an overview of the six editions of this challenge series since 2015, discuss the lessons learned, and outline the areas which need more work or give puzzling results.
Inducing semantic representations directly from speech signals is a highly challenging task but has many useful applications in speech mining and spoken language understanding. This study tackles the unsupervised learning of semantic representations for spoken utterances. Through converting speech signals into hidden units generated from acoustic unit discovery, we propose WavEmbed, a multimodal sequential autoencoder that predicts hidden units from a dense representation of speech. Secondly, we also propose S-HuBERT to induce meaning through knowledge distillation, in which a sentence embedding model is first trained on hidden units and passes its knowledge to a speech encoder through contrastive learning. The best performing model achieves a moderate correlation (0.5~0.6) with human judgments, without relying on any labels or transcriptions. Furthermore, these models can also be easily extended to leverage textual transcriptions of speech to learn much better speech embeddings that are strongly correlated with human annotations. Our proposed methods are applicable to the development of purely data-driven systems for speech mining, indexing and search.
ASR has been shown to achieve great performance recently. However, most of them rely on massive paired data, which is not feasible for low-resource languages worldwide. This paper investigates how to learn directly from unpaired phone sequences and speech utterances. We design a two-stage iterative framework. GAN training is adopted in the first stage to find the mapping relationship between unpaired speech and phone sequence. In the second stage, another HMM model is introduced to train from the generator's output, which boosts the performance and provides a better segmentation for the next iteration. In the experiment, we first investigate different choices of model designs. Then we compare the framework to different types of baselines: (i) supervised methods (ii) acoustic unit discovery based methods (iii) methods learning from unpaired data. Our framework performs consistently better than all acoustic unit discovery methods and previous methods learning from unpaired data based on the TIMIT dataset.
Latent Dirichlet allocation (LDA) is widely used for unsupervised topic modelling on sets of documents. No temporal information is used in the model. However, there is often a relationship between the corresponding topics of consecutive tokens. In this paper, we present an extension to LDA that uses a Markov chain to model temporal information. We use this new model for acoustic unit discovery from speech. As input tokens, the model takes a discretised encoding of speech from a vector quantised (VQ) neural network with 512 codes. The goal is then to map these 512 VQ codes to 50 phone-like units (topics) in order to more closely resemble true phones. In contrast to the base LDA, which only considers how VQ codes co-occur within utterances (documents), the Markov chain LDA additionally captures how consecutive codes follow one another. This extension leads to an increase in cluster quality and phone segmentation results compared to the base LDA. Compared to a recent vector quantised neural network approach that also learns 50 units, the extended LDA model performs better in phone segmentation but worse in mutual information.
The success of deep learning comes from its ability to capture the hierarchical structure of data by learning high-level representations defined in terms of low-level ones. In this paper we explore self-supervised learning of hierarchical representations of speech by applying multiple levels of Contrastive Predictive Coding (CPC). We observe that simply stacking two CPC models does not yield significant improvements over single-level architectures. Inspired by the fact that speech is often described as a sequence of discrete units unevenly distributed in time, we propose a model in which the output of a low-level CPC module is non-uniformly downsampled to directly minimize the loss of a high-level CPC module. The latter is designed to also enforce a prior of separability and discreteness in its representations by enforcing dissimilarity of successive high-level representations through focused negative sampling, and by quantization of the prediction targets. Accounting for the structure of the speech signal improves upon single-level CPC features and enhances the disentanglement of the learned representations, as measured by downstream speech recognition tasks, while resulting in a meaningful segmentation of the signal that closely resembles phone boundaries.
Recent work on unsupervised speech segmentation has used self-supervised models with phone and word segmentation modules that are trained jointly. This paper instead revisits an older approach to word segmentation: bottom-up phone-like unit discovery is performed first, and symbolic word segmentation is then performed on top of the discovered units (without influencing the lower level). To do this, I propose a new unit discovery model, a new symbolic word segmentation model, and then chain the two models to segment speech. Both models use dynamic programming to minimize segment costs from a self-supervised network with an additional duration penalty that encourages longer units. Concretely, for acoustic unit discovery, duration-penalized dynamic programming (DPDP) is used with a contrastive predictive coding model as the scoring network. For word segmentation, DPDP is applied with an autoencoding recurrent neural as the scoring network. The two models are chained in order to segment speech. This approach gives comparable word segmentation results to state-of-the-art joint self-supervised segmentation models on an English benchmark. On French, Mandarin, German and Wolof data, it outperforms previous systems on the ZeroSpeech benchmarks. Analysis shows that the chained DPDP system segments shorter filler words well, but longer words might require some external top-down signal.
Contrastive predictive coding (CPC) aims to learn representations of speech by distinguishing future observations from a set of negative examples. Previous work has shown that linear classifiers trained on CPC features can accurately predict speaker and phone labels. However, it is unclear how the features actually capture speaker and phonetic information, and whether it is possible to normalize out the irrelevant details (depending on the downstream task). In this paper, we first show that the per-utterance mean of CPC features captures speaker information to a large extent. Concretely, we find that comparing means performs well on a speaker verification task. Next, probing experiments show that standardizing the features effectively removes speaker information. Based on this observation, we propose a speaker normalization step to improve acoustic unit discovery using K-means clustering of CPC features. Finally, we show that a language model trained on the resulting units achieves some of the best results in the ZeroSpeech2021~Challenge.
Discovering speaker independent acoustic units purely from spoken input is known to be a hard problem. In this work we propose an unsupervised speaker normalization technique prior to unit discovery. It is based on separating speaker related from content induced variations in a speech signal with an adversarial contrastive predictive coding approach. This technique does neither require transcribed speech nor speaker labels, and, furthermore, can be trained in a multilingual fashion, thus achieving speaker normalization even if only few unlabeled data is available from the target language. The speaker normalization is done by mapping all utterances to a medoid style which is representative for the whole database. We demonstrate the effectiveness of the approach by conducting acoustic unit discovery with a hidden Markov model variational autoencoder noting, however, that the proposed speaker normalization can serve as a front end to any unit discovery system. Experiments on English, Yoruba and Mboshi show improvements compared to using non-normalized input.
This paper tackles automatically discovering phone-like acoustic units (AUD) from unlabeled speech data. Past studies usually proposed single-step approaches. We propose a two-stage approach: the first stage learns a subword-discriminative feature representation and the second stage applies clustering to the learned representation and obtains phone-like clusters as the discovered acoustic units. In the first stage, a recently proposed method in the task of unsupervised subword modeling is improved by replacing a monolingual out-of-domain (OOD) ASR system with a multilingual one to create a subword-discriminative representation that is more language-independent. In the second stage, segment-level k-means is adopted, and two methods to represent the variable-length speech segments as fixed-dimension feature vectors are compared. Experiments on a very low-resource Mboshi language corpus show that our approach outperforms state-of-the-art AUD in both normalized mutual information (NMI) and F-score. The multilingual ASR improved upon the monolingual ASR in providing OOD phone labels and in estimating the phone boundaries. A comparison of our systems with and without knowing the ground-truth phone boundaries showed a 16% NMI performance gap, suggesting that the current approach can significantly benefit from improved phone boundary estimation.
In this work, we propose a hierarchical subspace model for acoustic unit discovery. In this approach, we frame the task as one of learning embeddings on a low-dimensional phonetic subspace, and simultaneously specify the subspace itself as an embedding on a hyper-subspace. We train the hyper-subspace on a set of transcribed languages and transfer it to the target language. In the target language, we infer both the language and unit embeddings in an unsupervised manner, and in so doing, we simultaneously learn a subspace of units specific to that language and the units that dwell on it. We conduct our experiments on TIMIT and two low-resource languages: Mboshi and Yoruba. Results show that our model outperforms major acoustic unit discovery techniques, both in terms of clustering quality and segmentation accuracy.
In this paper, we explore vector quantization for acoustic unit discovery. Leveraging unlabelled data, we aim to learn discrete representations of speech that separate phonetic content from speaker-specific details. We propose two neural models to tackle this challenge - both use vector quantization to map continuous features to a finite set of codes. The first model is a type of vector-quantized variational autoencoder (VQ-VAE). The VQ-VAE encodes speech into a sequence of discrete units before reconstructing the audio waveform. Our second model combines vector quantization with contrastive predictive coding (VQ-CPC). The idea is to learn a representation of speech by predicting future acoustic units. We evaluate the models on English and Indonesian data for the ZeroSpeech 2020 challenge. In ABX phone discrimination tests, both models outperform all submissions to the 2019 and 2020 challenges, with a relative improvement of more than 30%. The models also perform competitively on a downstream voice conversion task. Of the two, VQ-CPC performs slightly better in general and is simpler and faster to train. Finally, probing experiments show that vector quantization is an effective bottleneck, forcing the models to discard speaker information.
For our submission to the ZeroSpeech 2019 challenge, we apply discrete latent-variable neural networks to unlabelled speech and use the discovered units for speech synthesis. Unsupervised discrete subword modelling could be useful for studies of phonetic category learning in infants or in low-resource speech technology requiring symbolic input. We use an autoencoder (AE) architecture with intermediate discretisation. We decouple acoustic unit discovery from speaker modelling by conditioning the AE's decoder on the training speaker identity. At test time, unit discovery is performed on speech from an unseen speaker, followed by unit decoding conditioned on a known target speaker to obtain reconstructed filterbanks. This output is fed to a neural vocoder to synthesise speech in the target speaker's voice. For discretisation, categorical variational autoencoders (CatVAEs), vector-quantised VAEs (VQ-VAEs) and straight-through estimation are compared at different compression levels on two languages. Our final model uses convolutional encoding, VQ-VAE discretisation, deconvolutional decoding and an FFTNet vocoder. We show that decoupled speaker conditioning intrinsically improves discrete acoustic representations, yielding competitive synthesis quality compared to the challenge baseline.
This work tackles the problem of learning a set of language specific acoustic units from unlabeled speech recordings given a set of labeled recordings from other languages. Our approach may be described by the following two steps procedure: first the model learns the notion of acoustic units from the labelled data and then the model uses its knowledge to find new acoustic units on the target language. We implement this process with the Bayesian Subspace Hidden Markov Model (SHMM), a model akin to the Subspace Gaussian Mixture Model (SGMM) where each low dimensional embedding represents an acoustic unit rather than just a HMM's state. The subspace is trained on 3 languages from the GlobalPhone corpus (German, Polish and Spanish) and the AUs are discovered on the TIMIT corpus. Results, measured in equivalent Phone Error Rate, show that this approach significantly outperforms previous HMM based acoustic units discovery systems and compares favorably with the Variational Auto Encoder-HMM.