Bayesian Subspace Hidden Markov Model for Acoustic Unit Discovery

Ondel, Lucas; Vydana, Hari Krishna; Burget, Lukáš; Černocký, Jaň

doi:10.21437/interspeech.2019-2224

Cited by 18 publications

(25 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The DNN AM in our system is trained and evaluated on the entire Mboshi dataset without training-test partition, as we are tackling an unsupervised learning problem. This is also consistent with the studies [11,19].…”

Section: Databasessupporting

confidence: 94%

“…There are two mainstream research strands in unsupervised acoustic modeling. The first strand, acoustic unit discovery (AUD) [6,10], formulates the problem as discovering a finite set of phone-like acoustic units [6,7,11]. The second strand, unsupervised subword modeling (USM) [9,12], formulates the problem as learning a frame-level feature representation that can Code: https://github.com/syfengcuhk/mboshi.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised Acoustic Unit Discovery by Leveraging a Language-Independent Subword Discriminative Feature Representation

et al. 2021

View full text Add to dashboard Cite

This paper tackles automatically discovering phone-like acoustic units (AUD) from unlabeled speech data. Past studies usually proposed single-step approaches. We propose a twostage 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 outof-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 kmeans is adopted, and two methods to represent the variablelength 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-theart 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.

show abstract

Section: Databasessupporting

confidence: 94%

Section: Introductionmentioning

confidence: 99%

Unsupervised Acoustic Unit Discovery by Leveraging a Language-Independent Subword Discriminative Feature Representation

et al. 2021

View full text Add to dashboard Cite

show abstract

“…This study is inspired by recent work showing the benefit of using multilingual bottleneck features as frame-level representations for zero-resource languages [23][24][25][26]. In [27], multilingual data were used in a similar way for discovering acoustic units. As in those studies, our findings show the advantage of learning from labelled data in well-resourced languages when processing an unseen low-resource languagehere at the word rather than subword level.…”

Section: Introductionmentioning

confidence: 99%

Multilingual Acoustic Word Embedding Models for Processing Zero-resource Languages

Kamper

Matusevych

Goldwater

2020

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Acoustic word embeddings are fixed-dimensional representations of variable-length speech segments. In settings where unlabelled speech is the only available resource, such embeddings can be used in "zero-resource" speech search, indexing and discovery systems. Here we propose to train a single supervised embedding model on labelled data from multiple well-resourced languages and then apply it to unseen zeroresource languages. For this transfer learning approach, we consider two multilingual recurrent neural network models: a discriminative classifier trained on the joint vocabularies of all training languages, and a correspondence autoencoder trained to reconstruct word pairs. We test these using a word discrimination task on six target zero-resource languages. When trained on seven well-resourced languages, both models perform similarly and outperform unsupervised models trained on the zero-resource languages. With just a single training language, the second model works better, but performance depends more on the particular training-testing language pair.Index Terms-Acoustic word embeddings, multilingual models, zero-resource speech processing, query-by-example.

show abstract

“…Vague priors are easy to define but they fail to provide a reasonable selection of "good" candidates. Recent works [17], [18] have proposed to remedy this shortcoming by introducing Fig. 2: Illustration of the concept of phonetic subspace: each phone is represented as a vector η encoding the parameters of a probabilistic model (an HMM in this example).…”

Section: B Phonetic Subspacementioning

confidence: 99%

Non-Parametric Bayesian Subspace Models for Acoustic Unit Discovery

Ondel¹

2021

Preprint

Self Cite

View full text Add to dashboard Cite

This work investigates subspace non-parametric models for the task of learning a set of acoustic units from unlabeled speech recordings. We constrain the base-measure of a Dirichlet-Process mixture with a phonetic subspace---estimated from other source languages---to build an \emph{educated prior}, thereby forcing the learned acoustic units to resemble phones of known source languages. Two types of models are proposed: (i) the Subspace HMM (SHMM) which assumes that the phonetic subspace is the same for every language, (ii) the Hierarchical-Subspace HMM (H-SHMM) which relaxes this assumption and allows to have a language-specific subspace estimated on the unlabeled target data. These models are applied on 3 languages: English, Yoruba and Mboshi and they are compared with various competitive acoustic units discovery baselines. Experimental results show that both subspace models outperform other systems in terms of clustering quality and segmentation accuracy. Moreover, we observe that the H-SHMM provides results superior to the SHMM supporting the idea that language-specific priors are preferable to language-agnostic priors for acoustic unit discovery.

show abstract

Bayesian Subspace Hidden Markov Model for Acoustic Unit Discovery

Cited by 18 publications

References 11 publications

Unsupervised Acoustic Unit Discovery by Leveraging a Language-Independent Subword Discriminative Feature Representation

Unsupervised Acoustic Unit Discovery by Leveraging a Language-Independent Subword Discriminative Feature Representation

Multilingual Acoustic Word Embedding Models for Processing Zero-resource Languages

Non-Parametric Bayesian Subspace Models for Acoustic Unit Discovery

Contact Info

Product

Resources

About