Untranscribed Web Audio for Low Resource Speech Recognition

Carmantini, Andrea; Bell, P. J.; Renals, Steve

doi:10.21437/interspeech.2019-2623

Cited by 13 publications

(18 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The original work was demonstrated on in-domain data-sets (conversational speech). Nevertheless, similar techniques have also been employed to use a seed model trained on out-of-domain data followed by SSL on in-domain data [1,17,18]. The latter is often obtained from data crawled from the web.…”

Section: Semi-supervised Training Using Lfmmimentioning

confidence: 99%

“…The acoustic model is then trained with the newly generated labels along with that of the supervised data. In this paper, we describe our efforts for the MATERIAL program 1 , where the training data consists of only conversational speech and the evaluation data consists of three genres: conversational speech, news and topical broadcast (CS, NB and TB, respectively). Moreover, unlike the training data, majority of test data belong to NB and TB.…”

Section: Introductionmentioning

confidence: 99%

“…This is both a multi-genre (in terms of speaking style and content) and a multi-bandwidth condition. Such domain adaptation problems can be addressed with SSL using data collected from the web [1][2][3].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Incremental Semi-Supervised Learning for Multi-Genre Speech Recognition

Khonglah

Madikeri

Dey

et al. 2020

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

View full text Add to dashboard Cite

In this work, we explore a data scheduling strategy for semisupervised learning (SSL) for acoustic modeling in automatic speech recognition. The conventional approach uses a seed model trained with supervised data to automatically recognize the entire set of unlabeled (auxiliary) data to generate new labels for subsequent acoustic model training. In this paper, we propose an approach in which the unlabelled set is divided into multiple equal-sized subsets. These subsets are processed in an incremental fashion: for each iteration a new subset is added to the data used for SSL, starting from only one subset in the first iteration. The acoustic model from the previous iteration becomes the seed model for the next one. This scheduling strategy is compared to the approach employing all unlabeled data in one-shot for training. Experiments using lattice-free maximum mutual information based acoustic model training on Fisher English gives 80% word error recovery rate. On the multi-genre evaluation sets on Lithuanian and Bulgarian relative improvements of up to 17.2% in word error rate are observed.

show abstract

Section: Semi-supervised Training Using Lfmmimentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Incremental Semi-Supervised Learning for Multi-Genre Speech Recognition

Khonglah

Madikeri

Dey

et al. 2020

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

View full text Add to dashboard Cite

show abstract

“…This semi-supervised approach to DNN training has been successfully used eg. [9,10,11]. However, the technique requires careful confidence-based data selection, and is very sensitive to the performance of the source system on the target data.…”

Section: Introductionmentioning

confidence: 99%

Cross Lingual Transfer Learning for Zero-Resource Domain Adaptation

Abad

Bell

Carmantini

et al. 2020

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

Self Cite

View full text Add to dashboard Cite

We propose a method for zero-resource domain adaptation of DNN acoustic models, for use in low-resource situations where the only in-language training data available may be poorly matched to the intended target domain. Our method uses a multi-lingual model in which several DNN layers are shared between languages. This architecture enables domain adaptation transforms learned for one well-resourced language to be applied to an entirely different lowresource language. First, to develop the technique we use English as a well-resourced language and take Spanish to mimic a low-resource language. Experiments in domain adaptation between the conversational telephone speech (CTS) domain and broadcast news (BN) domain demonstrate a 29% relative WER improvement on Spanish BN test data by using only English adaptation data. Second, we demonstrate the effectiveness of the method for low-resource languages with a poor match to the well-resourced language. Even in this scenario, the proposed method achieves relative WER improvements of 18-27% by using solely English data for domain adaptation. Compared to other related approaches based on multi-task and multi-condition training, the proposed method is able to better exploit well-resource language data for improved acoustic modelling of the low-resource target domain.

show abstract

“…LF-MMI trained AMs have also been effective in transfer learning based domain adaptation, wherein an ASR model trained on a large out-of-domain dataset is adapted in a supervised [7] or semi-supervised way [8] to a smaller in-domain dataset. Semisupervised LF-MMI based training has also been used for lowresourced languages [9]. Yet, these works have typically relied on hundreds of hours of in-domain data.…”

Section: Introductionmentioning

confidence: 99%

On Semi-Supervised LF-MMI Training of Acoustic Models with Limited Data

Sheikh¹,

Illina²

2020

Interspeech 2020

View full text Add to dashboard Cite

This work investigates semi-supervised training of acoustic models (AM) with the lattice-free maximum mutual information (LF-MMI) objective in practically relevant scenarios with a limited amount of labeled in-domain data. An error detection driven semi-supervised AM training approach is proposed, in which an error detector controls the hypothesized transcriptions or lattices used as LF-MMI training targets on additional unlabeled data. Under this approach, our first method uses a single error-tagged hypothesis whereas our second method uses a modified supervision lattice. These methods are evaluated and compared with existing semi-supervised AM training methods in three different matched or mismatched, limited data setups. Word error recovery rates of 28 to 89% are reported.

show abstract

Untranscribed Web Audio for Low Resource Speech Recognition

Cited by 13 publications

References 26 publications

Incremental Semi-Supervised Learning for Multi-Genre Speech Recognition

Incremental Semi-Supervised Learning for Multi-Genre Speech Recognition

Cross Lingual Transfer Learning for Zero-Resource Domain Adaptation

On Semi-Supervised LF-MMI Training of Acoustic Models with Limited Data

Contact Info

Product

Resources

About