Learning non-parametric models of pronunciation

Hutchinson, Brian; Droppo, Jasha

doi:10.1109/icassp.2011.5947455

Cited by 5 publications

(8 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our approach is similar, in spirit, to previous work on learning non-parametric pronunciation models [6], learning contextdependent string edit distances [7] and phone-to-word transduction [8]. The novelty of our approach, distinguishing it from previous methods [6,7], lies in the fact that we do not treat q w and q x simply as sequences of phonetic labels: we explicitly exploit timing information associated with phone-boundaries in these sequences, which may be easily obtained as part of the forced-alignment or phonetic decoding process.…”

Section: Introductionmentioning

confidence: 90%

“…The novelty of our approach, distinguishing it from previous methods [6,7], lies in the fact that we do not treat q w and q x simply as sequences of phonetic labels: we explicitly exploit timing information associated with phone-boundaries in these sequences, which may be easily obtained as part of the forced-alignment or phonetic decoding process. In addition, the proposed approach naturally incorporates long-span context since it is defined in terms of contiguous sequences of phonetic labels -which we term as chunks -similar to the multigram model of Deligne et al [9].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A chunk-based phonetic score for mobile voice search

Prabhavalkar

Droppo

2012

2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

We propose a chunk-based phonetic score for re-scoring word hypotheses for the mobile voice search task. The score is based on a novel technique for aligning decoded phone sequences with forced-alignments of hypothesized word sequences and exploits phone-boundary timing information. In experimental results, we find that the proposed approach results in relative a word error rate reduction of 4.4% and a relative sentence error rate reduction of 2.3% for the Windows live search for mobile task [1].

show abstract

Section: Introductionmentioning

confidence: 90%

Section: Introductionmentioning

confidence: 99%

A chunk-based phonetic score for mobile voice search

Prabhavalkar

Droppo

2012

2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…One approach, which was studied heavily especially in the 1990s but also more recently, is to start with a dictionary containing canonical pronunciations and add to it those alternative pronunciations that occur often in some database, or that are generated by deterministic or probabilistic phonetic substitution, insertion, and deletion rules (e.g., Sloboda and Waibel, 1996;Riley et al, 1999;Weintraub et al, 1996b;Strik and Cucchiarini, 1999;FoslerLussier, 1999;Saraçlar and Khudanpur, 2004;Hazen et al, 2005). Other approaches are based on alternative models of transformations between the canonical and observed pronunciations, such as phonetic edit distance models (Hutchinson and Droppo, 2011) and log-linear models with features based on canonical-observed phone string combinations (Zweig and Nguyen, 2009). Efforts to use such ideas in ASR systems have produced performance gains, but not of sufficient magnitude to solve the pronunciation variation problem.…”

Section: Asrmentioning

confidence: 97%

Articulatory feature-based pronunciation modeling

Livescu

Jyothi

Fosler‐Lussier

2016

Computer Speech & Language

View full text Add to dashboard Cite

Please cite this article in press as: Livescu, K., et al., Articulatory feature-based pronunciation modeling. Comput. Speech Lang. (2015), http://dx. AbstractSpoken language, especially conversational speech, is characterized by great variability in word pronunciation, including many variants that differ grossly from dictionary prototypes. This is one factor in the poor performance of automatic speech recognizers on conversational speech, and it has been very difficult to mitigate in traditional phone-based approaches to speech recognition. An alternative approach, which has been studied by ourselves and others, is one based on sub-phonetic features rather than phones. In such an approach, a word's pronunciation is represented as multiple streams of phonological features rather than a single stream of phones. Features may correspond to the positions of the speech articulators, such as the lips and tongue, or may be more abstract categories such as manner and place.This article reviews our work on a particular type of articulatory feature-based pronunciation model. The model allows for asynchrony between features, as well as per-feature substitutions, making it more natural to account for many pronunciation changes that are difficult to handle with phone-based models. Such models can be efficiently represented as dynamic Bayesian networks. The feature-based models improve significantly over phone-based counterparts in terms of frame perplexity and lexical access accuracy. The remainder of the article discusses related work and future directions.

show abstract

“…For example, phonetic dictionary expansion may affect different systems differently (e.g., possibly achieving greater improvements in a segment-based recognizer [33] than in HMM-based recognizers [30], [10]), but to our knowledge there have been no direct comparisons on identical tasks and data sets. We have also only briefly touched on automatic sub-word unit learning and the related task of automatic dictionary learning [39], [40], [47].…”

Section: Discussionmentioning

confidence: 99%

“…Dictionaries are typically manually generated, but can also be generated in a data-driven way [39], [40].…”

Section: B Impact Of Phonetic Dictionary Expansionmentioning

confidence: 99%

Subword Modeling for Automatic Speech Recognition: Past, Present, and Emerging Approaches

Livescu¹,

Fosler‐Lussier

Metze

2012

IEEE Signal Process. Mag.

View full text Add to dashboard Cite

Abstract-Modern automatic speech recognition systems handle large vocabularies of words, making it infeasible to collect enough repetitions of each word to train individual word models. Instead, large-vocabulary recognizers represent each word in terms of sub-word units. Typically the sub-word unit is the phone, a basic speech sound such as a single consonant or vowel. Each word is then represented as a sequence, or several alternative sequences, of phones specified in a pronunciation dictionary. Other choices of sub-word units have been studied as well. The choice of sub-word units, and the way in which the recognizer represents words in terms of combinations of those units, is the problem of sub-word modeling. Different sub-word models may be preferable in different settings, such as high-variability conversational speech, high-noise conditions, low-resource settings, or multilingual speech recognition. This article reviews past, present, and emerging approaches to sub-word modeling. In order to make clean comparisons between many approaches, the review uses the unifying language of graphical models.

show abstract

Learning non-parametric models of pronunciation

Cited by 5 publications

References 18 publications

A chunk-based phonetic score for mobile voice search

A chunk-based phonetic score for mobile voice search

Articulatory feature-based pronunciation modeling

Subword Modeling for Automatic Speech Recognition: Past, Present, and Emerging Approaches

Contact Info

Product

Resources

About