Combination of strongly and weakly constrained recognizers for reliable detection of OOVS

Burget, Lukáš; Schwarz, Petr; Matějka, Pavel; Hannemann, Mirko; Rastrow, Ariya; White, C.; Khudanpur, Sanjeev; Heřmanský, Hynek; Černocký, Jaň

doi:10.1109/icassp.2008.4518551

Cited by 45 publications

(34 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our proposed selection technique we use a word confidence measure Cmax based on these frame level word posteriors [38], given as the maximum word confidence of the word in its hypothesized time interval (ts, t f )…”

Section: Asr-based Word Confidence Scoresmentioning

confidence: 99%

Deep neural network features and semi-supervised training for low resource speech recognition

Thomas

Seltzer

Church³

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

138

100

View full text Add to dashboard Cite

We propose a new technique for training deep neural networks (DNNs) as data-driven feature front-ends for large vocabulary continuous speech recognition (LVCSR) in low resource settings. To circumvent the lack of sufficient training data for acoustic modeling in these scenarios, we use transcribed multilingual data and semi-supervised training to build the proposed feature front-ends. In our experiments, the proposed features provide an absolute improvement of 16% in a low-resource LVCSR setting with only one hour of in-domain training data. While close to three-fourths of these gains come from DNN-based features, the remaining are from semi-supervised training.

show abstract

Section: Asr-based Word Confidence Scoresmentioning

confidence: 99%

Deep neural network features and semi-supervised training for low resource speech recognition

Thomas

Seltzer

Church³

et al. 2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

138

100

View full text Add to dashboard Cite

show abstract

“…It is clear from Fig. 1 that the condence measurement based on fragment posterior probability (dotdashed line tagged as P f rag ) in hybrid confusion networks outperforms existing methods based on condence measures from LVCSR systems and word entropy (solid line further from the origin) of the word based system [4]. As shown by the closest solid line to origin in Fig.…”

Section: Discussionmentioning

confidence: 68%

“…Although sub word posteriors are very good for detecting OOV regions, we also explored the use of additional features in Eqn. 2 that contain complimentary information such as those used in the JHU workshop [4]. They include:…”

Section: Fragment Posteriors Using Consensusmentioning

confidence: 99%

“…Examples of such approaches can be found in [1,2,3] 2. Confidence Scores More recent approaches are focused on detecting OOVs based on some condence measures such as acoustic scores, statistics derived from the language model and statistics derived from N-best lists (or lattices) [4,5,6,7] In general, methods using condence scores have a better performance for OOV detection. The main weakness of this strategy is that such condence measures are good at predicting whether the hypothesized word is correct or not, but unable to tease apart errors due to OOV words from those errors due to other phenomena such as degraded acoustic conditions.…”

Section: Introductionmentioning

confidence: 99%

“…The main weakness of this strategy is that such condence measures are good at predicting whether the hypothesized word is correct or not, but unable to tease apart errors due to OOV words from those errors due to other phenomena such as degraded acoustic conditions. In [4,7] the word-based system (strongly constrained with a language model) and the phone recognizer are used in parallel to address the OOV detection problem.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A new method for OOV detection using hybrid word/fragment system

Rastrow

Sethy

Ramabhadran

2009

2009 IEEE International Conference on Acoustics, Speech and Signal Processing

Self Cite

View full text Add to dashboard Cite

In this paper, we propose a new method for detecting regions with out-of-vocabulary (OOV) words in the output of a large vocabulary continuous speech recognition (LVCSR) system. The proposed method uses a hybrid system combining words and data-driven variable length sub word units. With the use of a single feature, the posterior probability of sub word units, this method outperforms existing methods published in the literature. We also presents a recipe to discriminatively train a hybrid language model to improve OOV detection rate. Results are presented on the RT04 broadcast news task.

show abstract

Dealing with Newly Emerging OOVs in Broadcast Programs by Daily Updates of the Lexicon and Language Model

Červa

Volna

Weingartová

2020

Speech and Computer

View full text Add to dashboard Cite

Combination of strongly and weakly constrained recognizers for reliable detection of OOVS

Cited by 45 publications

References 6 publications

Deep neural network features and semi-supervised training for low resource speech recognition

Deep neural network features and semi-supervised training for low resource speech recognition

A new method for OOV detection using hybrid word/fragment system

Dealing with Newly Emerging OOVs in Broadcast Programs by Daily Updates of the Lexicon and Language Model

Contact Info

Product

Resources

About