Advances in transcription of broadcast news and conversational telephone speech within the combined EARS BBN/LIMSI system

Matsoukas, Spyros; Gauvain, Jean‐Luc; Adda, Gilles; Colthurst, Thomas; Kao, Chia-Lin; Kimball, Owen; Lamel, Lori; Lefèvre, Fabrice; Ma, Jiyong; Makhoul, John; Nguyen, Long; Prasad, Rohit; Schwartz, Richard; Schwenk, Holger; Xiang, Bing

doi:10.1109/tasl.2006.878257

Cited by 40 publications

(15 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unsupervised AM adaptation is performed for each segment cluster using the CMLLR and MLLR techniques and relies on a tree organization of the tied states to create the regression classes as a function of the available data. Different combinations of automatic segmentations (GMM or BIC based), acoustic models and language models are used in the different passes [10]. The acoustic models are MLLT-SAT trained on 1044 hours of data.…”

Section: Limsimentioning

confidence: 99%

Speech Recognition System Combination for Machine Translation

Gales

Liu

Sinha

et al. 2007

2007 IEEE International Conference on Acoustics, Speech and Signal Processing - ICASSP '07

Self Cite

View full text Add to dashboard Cite

The majority of state-of-the-art speech recognition systems make use of system combination. The combination approaches adopted have traditionally been tuned to minimising Word Error Rates (WERs). In recent years there has been growing interest in taking the output from speech recognition systems in one language and translating it into another. This paper investigates the use of cross-site combination approaches in terms of both WER and impact on translation performance. In addition the stages involved in modifying the output from a Speech-to-Text (STT) system to be suitable for translation are described. Two source languages, Mandarin and Arabic, are recognised and then translated using a phrase-based statistical machine translation system into English. Performance of individual systems and cross-site combination using cross-adaptation and ROVER are given. Results show that the best STT combination scheme in terms of WER is not necessarily the most appropriate when translating speech.

show abstract

Section: Limsimentioning

confidence: 99%

Speech Recognition System Combination for Machine Translation

Gales

Liu

Sinha

et al. 2007

2007 IEEE International Conference on Acoustics, Speech and Signal Processing - ICASSP '07

Self Cite

View full text Add to dashboard Cite

show abstract

“…The primary reason has been the hope that human neural network like models may ultimately lead to human like performance. Early attempts at using ANNs for speech recognition centred on simple tasks like recognising a few phonemes or a few words or isolated digits, with satisfactory success (Lippmann, 1990;Evermann et al, 2004;Matsoukas et al, 2006), using pattern mapping by multi layer perceptron (MLP). However, limited ability of ANNs to capture temporal information from the speech signal was a major drawback.…”

Section: Sarmamentioning

confidence: 99%

Speech recognition using deep neural network - recent trends

Sarma

2017

IJISDC

View full text Add to dashboard Cite

Deep neural networks (DNN) are special forms of learning-based structures composed of multiple hidden layers formed by artificial neurons. These are different to the conventional artificial neural networks (ANN) and are accepted as efficient tools for solving emerging real world problems. Recently, DNNs have become a mainstream speech recognition tool and are fast becoming part of evolving technologies emerging as a viable option to replace all other leading tools so far used. ANNs with deep learning which uses a generative, layer by-layer pre-training method for initialising the weights has provided best solution for acoustic modelling for speech recognition. This paper provides a brief description of the current technology related to speech recognition and its slow adoption of DNN-based approaches. Initially, a historical note on the technology development for speech recognition system is given. The later part explains the DNN-based acoustic modelling for speech recognition and recent technology developments reported and the ones available for actual use.Keywords: speech recognition; artificial neural network; ANN; deep neural network; DNN.Reference to this paper should be made as follows: Sarma, M. (2017) 'Speech recognition using deep neural network -recent trends', Int.

show abstract

“…The brains impressive superiority at a wide range of cognitive skills like speech recognition, has motivated the researchers to explore the possibilities of ANN models in the field of speech recognition in 1980s [24], with a hope that human neural network like models may ultimately lead to human-like performance. Early attempts at using neural networks for speech recognition centered on simple tasks like recognizing a few phonemes or a few words or isolated digits, with good success [25][26][27], using pattern mapping by multilayer perceptron (MLP). But at the later half of 1990, suddenly ANN-based speech research got terminated [24] after the statistical framework HMM come into focus, which supports both acoustic and temporal modeling of speech.…”

Section: Early Speech Recognition Technologymentioning

confidence: 99%

Speech Recognition in Indian Languages—A Survey

Sarma

2015

Recent Trends in Intelligent and Emerging Systems

View full text Add to dashboard Cite

In this paper, a brief overview derived out of detailed survey of speech recognition works reported in Indian languages is described. Robustness of speech recognition systems toward language variation is the recent trend of research in speech recognition technology. To develop a system which can communicate with human in any language like any other human is the foremost requirement in order to design appropriate speech recognition technology for one to all. India is a country which has vast linguistic variations among its billion plus population. Therefore, it provides a sound area of research toward language-specific speech recognition technology. From the beginning of the commercial availability of the speech recognition system, the technology has been dominated by the hidden Markov model (HMM) methodology due to its capability of modeling temporal structures of speech and encoding them as a sequence of spectral vectors. Most of the work done in Indian languages also uses HMM technology. However, from the last 10-15 years after the acceptance of neurocomputing as an alternative to HMM, artificial neural network (ANN)-based methodologies have started to receive attention for application in speech recognition. This is a trend worldwide as part of which few works have also been reported by a few researchers.

show abstract

Advances in transcription of broadcast news and conversational telephone speech within the combined EARS BBN/LIMSI system

Cited by 40 publications

References 23 publications

Speech Recognition System Combination for Machine Translation

Speech Recognition System Combination for Machine Translation

Speech recognition using deep neural network - recent trends

Speech Recognition in Indian Languages—A Survey

Contact Info

Product

Resources

About