Global optimization of a neural network-hidden Markov model hybrid

Bengio, Yoshua; Mori, Renato De; Flammia, Giovanni; Kompe, Ralf

doi:10.1109/72.125866

Cited by 154 publications

(29 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Since it is not known how t o compute this probability directly 4 , (1) is usually used to split this posterior probability i n to a likelihood p(X jM ) that represents the contribution of the acoustic model, and a prior probability P (Mj ) that represents the contribution of the language model. p(X jM) and P (M) are estimated during recognition from subsystems whose parameters are sometimes trained from di erent training sets (for instance, from an acoustic training set and from a large corpus of written text).…”

Section: Hidden Markov Models (Hmms)mentioning

confidence: 99%

Hybrid HMM/ANN systems for speech recognition: Overview and new research directions

Bourlard

Morgan

1998

Adaptive Processing of Sequences and Data Structures

View full text Add to dashboard Cite

Section: Hidden Markov Models (Hmms)mentioning

confidence: 99%

Hybrid HMM/ANN systems for speech recognition: Overview and new research directions

Bourlard

Morgan

1998

Adaptive Processing of Sequences and Data Structures

View full text Add to dashboard Cite

“…The direct optimization of such a model (Bengio, de Mori, Flammia & Kompe, 1992) is a computationally expensive process. However, using similar factorizations and assumptions as above, Bourlard, Konig and Morgan (1996) and Hennebert, Ris, Bourlard, Renals and Morgan (1997) demonstrated that a generalized EM algorithm exists for the optimization of the parameters of acceptor HMMs.…”

Section: Generative Hmms and Acceptor Hmmsmentioning

confidence: 99%

Confidence measures from local posterior probability estimates

Williams

Renals

1999

Computer Speech & Language

View full text Add to dashboard Cite

In this paper we introduce a set of related confidence measures for large vocabulary continuous speech recognition (LVCSR) based on local phone posterior probability estimates output by an acceptor HMM acoustic model. In addition to their computational efficiency, these confidence measures are attractive as they may be applied at the state-, phone-, word-or utterance-levels, potentially enabling discrimination between different causes of low confidence recognizer output, such as unclear acoustics or mismatched pronunciation models. We have evaluated these confidence measures for utterance verification using a number of different metrics. Experiments reveal several trends in "profitability of rejection", as measured by the unconditional error rate of a hypothesis test. These trends suggest that crude pronunciation models can mask the relatively subtle reductions in confidence caused by out-of-vocabulary (OOV) words and disfluencies, but not the gross model mismatches elicited by non-speech sounds. The observation that a purely acoustic confidence measure can provide improved performance over a measure based upon both acoustic and language model information for data drawn from the Broadcast News corpus, but not for data drawn from the North American Business News corpus suggests that the quality of model fit offered by a trigram language model is reduced for Broadcast News data. We also argue that acoustic confidence measures may be used to inform the search for improved pronunciation models.

show abstract

“…One solution uses the ANN to compute an additional set of symbols as transformed observations for the HMM [90]. A further improvement of this method is achieved through a global optimization of both the ANN and HMM [91]. This approach uses the gradient of the HMM optimization criterion with respect to the transformed observations to estimate the weights of the ANN connections.…”

Section: Proposed Solutionsmentioning

confidence: 99%