BUT OpenSAT 2017 Speech Recognition System

Karafiát, Martin; Baskar, Murali Karthick; Szöke, Igor; Malenovsky, Vladimir; Veselý, Karel; Grézl, František; Burget, Lukáš; Černocký, Jaň

doi:10.21437/interspeech.2018-2457

Cited by 4 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We can see that the overall behavior of individual systems stays similar and the input 0-gram augmentation brings clearly the largest gain. During the work on this paper, we have applied the 0-gram input augmentation to speech recognition in the OpenSAT challenge [15]. This task is considerable easier, with WER at around 10 %.…”

Section: Chime-6 Rescoring Resultsmentioning

confidence: 99%

Text Augmentation for Language Models in High Error Recognition Scenario

Beneš¹,

Burget²

2021

Interspeech 2021

Self Cite

View full text Add to dashboard Cite

We examine the effect of data augmentation for training of language models for speech recognition. We compare augmentation based on global error statistics with one based on per-word unigram statistics of ASR errors and observe that it is better to only pay attention the global substitution, deletion and insertion rates. This simple scheme also performs consistently better than label smoothing and its sampled variants. Additionally, we investigate into the behavior of perplexity estimated on augmented data, but conclude that it gives no better prediction of the final error rate. Our best augmentation scheme increases the absolute WER improvement from second-pass rescoring from 1.1 % to 1.9 % absolute on the CHiMe-6 challenge.

show abstract

Section: Chime-6 Rescoring Resultsmentioning

confidence: 99%

Text Augmentation for Language Models in High Error Recognition Scenario

Beneš¹,

Burget²

2021

Interspeech 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…Graves [17] first tried to use BLSTM for acoustic modeling of speech recognition and achieved the best recognition performance at that time on the TIMIT corpus. Subsequently, many researchers have studied the lowresource speech acoustic modeling of BLSTM [18]- [21]. Although the BLSTM has achieved good results, the structure has a large number of parameters and some complex training mechanisms.…”

Section: ) Recurrent Neural Networkmentioning

confidence: 99%

Acoustic Modeling Based on Deep Learning for Low-Resource Speech Recognition: An Overview

Kang

Chen

et al. 2020

IEEE Access

View full text Add to dashboard Cite

The polarization of world languages is becoming more and more obvious. Many languages, mainly endangered languages, are of low-resource attribute due to lack of information. Both language conservation and cultural heritage face important challenges. Therefore, speech recognition for lowresource scenario has become a hot topic in the field of speech. Based on the complex network structures and huge model parameters, deep learning has become a powerful science in the process of speech recognition, which has a broad and far-reaching significance for the study of low-resource speech recognition. Aiming at the characteristic of low resource, this paper reviews the history and research status of two kinds of acoustic models of deep learning neural networks and acoustic end-to-end structures. We further elaborate on several key techniques for improving performance in the two aspects of data and model training. There are two projects for low-resource languages introduced in this paper. The possible future developments are finally pointed out. These works provide some reference for computer speech and language processing.

show abstract

“…On contrary to the pilot OpenSAT evaluations conducted in 2017 [5], where the target data was real fireman-dispatcher communication from the Charleston Sofa Super Store Fire in 2007, for Open-SAT 2019, NIST prepared simulated public safety communications collection ("SAFE-T") specifically designed for speech analytic systems. The data is intended to simulate a combination of characteristics found in public safety communications: background noises, channel transmission noises and speaking characteristics such as stress or sense of urgency.…”

Section: Public Safety Communicationsmentioning

confidence: 99%

“…Various classical hybrid DNN-HMM (Hidden Markov Model) speech recognizers were trained in Kaldi toolkit [6]. We decided to select Factorized Time Delay NN (TDNN-F) based architectures [7] as we were consistently obtaining superior performance to recurrent NN architectures used in last OpenSAT evaluations [5]:…”

Section: Hybrid Acoustics Modelsmentioning

confidence: 99%

“…In previous evaluations [4,5] we were using block-softmax objective function [11] which is not yet implemented with LF-MMI training in Kaldi. Therefore, we chose a much simpler approach: The training language data (audio lists, transcriptions, pronunciation dictionaries...) are simply stacked together with added language indicator for phoneme names as suggested in [12].…”

Section: Multilingual Hybrid Modelsmentioning

confidence: 99%

See 1 more Smart Citation

BUT Opensat 2019 Speech Recognition System

Karafiát¹,

Baskar²,

Szöke³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

The paper describes the BUT Automatic Speech Recognition (ASR) systems submitted for OpenSAT evaluations under two domain categories such as low resourced languages and public safety communications. The first was challenging due to lack of training data, therefore various architectures and multilingual approaches were employed. The combination led to superior performance. The second domain was challenging due to recording in extreme conditions such as specific channel, speaker under stress and high levels of noise. Data augmentation process was inevitable to get reasonably good performance.

show abstract

BUT OpenSAT 2017 Speech Recognition System

Cited by 4 publications

References 0 publications

Text Augmentation for Language Models in High Error Recognition Scenario

Text Augmentation for Language Models in High Error Recognition Scenario

Acoustic Modeling Based on Deep Learning for Low-Resource Speech Recognition: An Overview

BUT Opensat 2019 Speech Recognition System

Contact Info

Product

Resources

About