BUT System for Low Resource Indian Language ASR

Pulugundla, Bhargav; Baskar, Murali Karthick; Kesiraju, Santosh; Egorova, Ekaterina; Karafiát, Martin; Burget, Lukáš; Černocký, Jaň

doi:10.21437/interspeech.2018-1302

Cited by 26 publications

(12 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The best ASR system, trained with the BSRD v2 dataset and using a 15-gram character-based language model, achieved a CER and a WER of 10.49% and 25.45%, respectively. This is similar to the result found in [9] (15.2% of WER) for Tamil language, and comparable to the improvement reported by [13] and [14] using a multilingual setting training and multimodal data augmentation, respectively. From these tables, one clearly verifies the advantage of incorporating the language model onto the acoustic models obtained in Section VIII, with the WER dropping from 71.62% to 30.50% for the BRSD v1 set, and a 21.96% improvement for the BRSD v2 set.…”

Section: Complete Asr Experimentssupporting

confidence: 89%

“…Most languages in the world lack the amount of text, speech, and/or linguistic resources required to build large models based on deep neural networks. There has been an increasing research interest on how to build a high-accuracy ASR system for languages with insufficient annotated data (using from hours to a few dozen hours of annotated speech), such as Brazilian Portuguese [8]; Indian languages (Gujarati, Tamil, and Telugu) [9], [10]; and Seneca (an Indigenous North-American language) [11].…”

Section: Development Context and General Diagram Of The Systemmentioning

confidence: 99%

See 1 more Smart Citation

An open-source end-to-end ASR system for Brazilian Portuguese using DNNs built from newly assembled corpora

Quintanilha

Netto

Biscainho³

2020

JCIS

View full text Add to dashboard Cite

In this work, we present a baseline end-to-end system based on deep learning for automatic speech recognition in Brazilian Portuguese. To build such a model, we employ a speech corpus containing 158 hours of annotated speech by assembling four individual datasets, three of them publicly available, and a text corpus containing 10.2 millions of sentences. We train an acoustic model based on the DeepSpeech 2 network, with two convolutional and five bidirectional recurrent layers. By adding a newly trained 15-gram language model at the character level, we achieve a character error rate of only 10.49% and a word error rate of 25.45%, which are on a par with other works in different languages using a similar amount of training data.

show abstract

Section: Complete Asr Experimentssupporting

confidence: 89%

Section: Development Context and General Diagram Of The Systemmentioning

confidence: 99%

An open-source end-to-end ASR system for Brazilian Portuguese using DNNs built from newly assembled corpora

Quintanilha

Netto

Biscainho³

2020

JCIS

View full text Add to dashboard Cite

show abstract

“…The Jilebi system [11] which performed the best across all languages was a multilingual TDNN based system with transfer learning. The system used an n-gram based LM for decoding and an RNN-based LM for rescoring.…”

Section: Resultsmentioning

confidence: 99%

Interspeech 2018 Low Resource Automatic Speech Recognition Challenge for Indian Languages

Srivastava¹,

Sitaram²,

Mehta³

et al. 2018

6th Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU 2018)

View full text Add to dashboard Cite

India has more than 1500 1 languages, with 30 of them spoken by more than one million native speakers. Most of them are low-resource and could greatly benefit from speech and language technologies. Building speech recognition support for these low-resource languages requires innovation in handling constraints on data size, while also exploiting the unique properties and similarities among Indian languages. With this goal, we organized a low-resource Automatic Speech Recognition challenge for Indian languages as part of Interspeech 2018. We released 50 hours of speech data with transcriptions for Tamil, Telugu and Gujarati, amounting to a total of 150 hours. Participants were required to only use the data we released for the challenge to preserve the low-resource setting, however, they were not restricted to work on any particular aspect of the speech recognizer. We received 109 submissions from 18 research groups and evaluated the systems in terms of Word Error Rate on a blind test set. In this paper we summarize the data, approaches and results of the challenge.

show abstract

“…This indicates that developing ASR systems for Indic languages with their richer phoneme/character sets and vocabulary is more challenging than English. Jilebi (Pulugundla et al, 2018) 14.0 13.9 14.7 Cogknit (Fathima et al, 2018) 17.7 16.0 17.1 CSALT-LEAP (Srivastava et al, 2018) -16.3 17.6 ISI-Billa (Billa, 2018) 19.3 19.6 20.9 MTL-SOL (Sailor and Hain, 2020) 18.4 16.3 18.6 Reed (Sen et al, 2021) 16.1 19.9 20.2 CNN + Context temporal features (Sen et al, 2020) 18…”

Section: Ablation Studies On Fine-tuning and Decodingmentioning

confidence: 99%

Towards Building ASR Systems for the Next Billion Users

Javed¹,

Doddapaneni²,

Raman³

et al. 2021

Preprint

View full text Add to dashboard Cite

Recent methods in speech and language technology pretrain very LARGE models which are fine-tuned for specific tasks. However, the benefits of such LARGE models are often limited to a few resource rich languages of the world. In this work, we make multiple contributions towards building ASR systems for low resource languages from the Indian subcontinent. First, we curate 17,000 hours of raw speech data for 40 Indian languages from a wide variety of domains including education, news, technology, and finance. Second, using this raw speech data we pretrain several variants of wav2vec style models for 40 Indian languages. Third, we analyze the pretrained models to find key features: codebook vectors of similar sounding phonemes are shared across languages, representations across layers are discriminative of the language family, and attention heads often pay attention within small local windows. Fourth, we fine-tune this model for downstream ASR for 9 languages and obtain state-of-the-art results on 3 public datasets, including on very low-resource languages such as Sinhala and Nepali. Our work establishes that multilingual pretraining is an effective strategy for building ASR systems for the linguistically diverse speakers of the Indian subcontinent.

show abstract

BUT System for Low Resource Indian Language ASR

Cited by 26 publications

References 14 publications

An open-source end-to-end ASR system for Brazilian Portuguese using DNNs built from newly assembled corpora

An open-source end-to-end ASR system for Brazilian Portuguese using DNNs built from newly assembled corpora

Interspeech 2018 Low Resource Automatic Speech Recognition Challenge for Indian Languages

Towards Building ASR Systems for the Next Billion Users

Contact Info

Product

Resources

About