Bayesian phonotactic Language Model for Acoustic Unit Discovery

Ondel, Lucas; Burget, Lukáš; Černocký, Jaň; Kesiraju, Santosh

doi:10.1109/icassp.2017.7953258

Cited by 8 publications

(20 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Surprisingly, the MBN features perform relatively poorly compared to the standard MFCC. These results are contradictory to those reported in [4]. Two factors may explain this discrepancy: the Mboshi5k data being different from the training data of the MBN neural network, the neural network may not generalize well.…”

Section: Resultscontrasting

confidence: 79%

“…If no match is found with a true phone boundary, this is considered to be a false negative. The consistency of the units was evaluated in term of normalized mutual information (NMI -see [2,4,6] for details) which measures the statistical dependency between the units and the forced aligned phones. A NMI of 0 % means that the units are completely independent of the phones whereas a NMI of 100 % indicates that the actual phones could be retrieved without error given the sequence of discovered units.…”

Section: Acoustic Unit Discovery (Aud) Evaluationmentioning

confidence: 99%

“…Building on the work of [18,19] they combine a nested hierarchical Pitman-Yor language model with a Weighted Finite State Transducer approach. Both for lattices and acoustic units sequences, we use the implementation of the authors with a bigram language model and a unigram character model 4 . Word discovery is evaluated using the Boundary metric from the Zero Resource Challenge 2017 [21] and [22].…”

Section: Extension To Word Discoverymentioning

confidence: 99%

“…This model, while rather crude, has shown that it can provide a clustering accurate O. Scharenborg was supported by a Vidi-grant from NWO (grant number: 276-89-003) enough to be used in topic identification of spoken document in unknown languages [3]. It was also shown that this model can be further improved by incorporating a Bayesian "phonotactic" language model learned jointly with the acoustic units [4]. Finally, following the work in [5] it has been combined successfully with variational auto-encoders leading to a model combining the potential of both deep neural networks and Bayesian models [6].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Bayesian Models for Unit Discovery on a Very Low Resource Language

Ondel

Godard

Besacier

et al. 2018

2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

Developing speech technologies for low-resource languages has become a very active research field over the last decade. Among others, Bayesian models have shown some promising results on artificial examples but still lack of in situ experiments. Our work applies state-of-the-art Bayesian models to unsupervised Acoustic Unit Discovery (AUD) in a real low-resource language scenario. We also show that Bayesian models can naturally integrate information from other resourceful languages by means of informative prior leading to more consistent discovered units. Finally, discovered acoustic units are used, either as the 1-best sequence or as a lattice, to perform word segmentation. Word segmentation results show that this Bayesian approach clearly outperforms a Segmental-DTW baseline on the same corpus.

show abstract

Section: Resultscontrasting

confidence: 79%

Section: Acoustic Unit Discovery (Aud) Evaluationmentioning

confidence: 99%

Section: Extension To Word Discoverymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Bayesian Models for Unit Discovery on a Very Low Resource Language

Ondel

Godard

Besacier

et al. 2018

2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…In addition, such methods may shed light on how human infants acquire language [2,3]. 1 https://github.com/kamperh/recipe_zs2017_track2 Several zero-resource tasks have been studied, including acoustic unit discovery [4][5][6], unsupervised representation learning [7][8][9], query-by-example search [10,11] and topic modelling [12,13]. Early work mainly focused on unsupervised term discovery, where the aim is to automatically find repeated word-or phrase-like patterns in a collection of speech [14][15][16].…”

Section: Introductionmentioning

confidence: 99%

An embedded segmental K-means model for unsupervised segmentation and clustering of speech

Kamper

Livescu

Goldwater

2017

2017 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)

View full text Add to dashboard Cite

Unsupervised segmentation and clustering of unlabelled speech are core problems in zero-resource speech processing. Most approaches lie at methodological extremes: some use probabilistic Bayesian models with convergence guarantees, while others opt for more efficient heuristic techniques. Despite competitive performance in previous work, the full Bayesian approach is difficult to scale to large speech corpora. We introduce an approximation to a recent Bayesian model that still has a clear objective function but improves efficiency by using hard clustering and segmentation rather than full Bayesian inference. Like its Bayesian counterpart, this embedded segmental K-means model (ES-KMeans) represents arbitrary-length word segments as fixed-dimensional acoustic word embeddings. We first compare ES-KMeans to previous approaches on common English and Xitsonga data sets (5 and 2.5 hours of speech): ES-KMeans outperforms a leading heuristic method in word segmentation, giving similar scores to the Bayesian model while being 5 times faster with fewer hyperparameters. However, its clusters are less pure than those of the other models. We then show that ES-KMeans scales to larger corpora by applying it to the 5 languages of the Zero Resource Speech Challenge 2017 (up to 45 hours), where it performs competitively compared to the challenge baseline. 1

show abstract

A Hierarchical Subspace Model for Language-Attuned Acoustic Unit Discovery

Yusuf

Ondel

Burget

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

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 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.

show abstract

Bayesian phonotactic Language Model for Acoustic Unit Discovery

Cited by 8 publications

References 8 publications

Bayesian Models for Unit Discovery on a Very Low Resource Language

Bayesian Models for Unit Discovery on a Very Low Resource Language

An embedded segmental K-means model for unsupervised segmentation and clustering of speech

A Hierarchical Subspace Model for Language-Attuned Acoustic Unit Discovery

Contact Info

Product

Resources

About