## Calibration of machine scores for pronunciation grading (1998)

Venue: | Proc. Int'l Conf. on Spoken Language Processing |

Citations: | 5 - 2 self |

### BibTeX

@INPROCEEDINGS{Franco98calibrationof,

author = {Horacio Franco and Leonardo Neumeyer},

title = {Calibration of machine scores for pronunciation grading},

booktitle = {Proc. Int'l Conf. on Spoken Language Processing},

year = {1998}

}

### OpenURL

### Abstract

Our proposed paradigm for automatic assessment of pronunciation quality uses hidden Markov models (HMMs) to generate phonetic segmentations of the student’s speech. From these segmentations, we use the HMMs to obtain spectral match and duration scores. In this work we focus on the problem of calibrating different machine scores to obtain an accurate prediction of the grades that a human expert would assign to the pronunciation. We discuss the application of different approaches based on minimum mean square error (MMSE) estimation and Bayesian classification. We investigate the characteristics of the different mappings as well as the effects of the prior distribution of grades in the calibration database. We finally suggest a simple method to extrapolate mappings from one language to another. 1.

### Citations

282 |
Neural network classifiers estimate Bayesian a posterior probabilities
- Richard, Lippmann
- 1991
(Show Context)
Citation Context ...e the input. Using standard training procedures, like backpropagation, neural networks will closely approximate Eq. (2), that is, the conditional expected value of the desired output given the inputs =-=[6]-=-, om ( ) ≅ Ehm [ ]. (6) Alternatively, if the network has G outputs corresponding to the grade classes, the network outputs will approximate the posterior probabilities Ph ( im) needed for classificat... |

50 | Automatic text-independent pronunciation scoring of foreign language student speech
- Neumeyer, Franco, et al.
- 1996
(Show Context)
Citation Context ...stimation problem, where we try to estimate, or predict, the value of the human grade by using a set of predictors. These predictors are the machine scores that we have presented in our previous work =-=[2]-=-,[3],[5]. We investigate the use of MMSE estimation and classification methods to predict the human grade from a set of machine scores. We present alternative implementations of these methods based on... |

42 | et al, “Automatic Pronunciation Scoring for Language Instruction
- Franco
(Show Context)
Citation Context ...ation problem, where we try to estimate, or predict, the value of the human grade by using a set of predictors. These predictors are the machine scores that we have presented in our previous work [2],=-=[3]-=-,[5]. We investigate the use of MMSE estimation and classification methods to predict the human grade from a set of machine scores. We present alternative implementations of these methods based on non... |

38 |
Genones: Optimizing the degree of mixture tying in a large vocabulary hidden Markov model based speech recognizer
- Digalakis, Murveit
- 1994
(Show Context)
Citation Context ...ruction; we address the task of grading the pronunciation quality of the speech of a student of a foreign language. The automatic grading system uses an HMM-based continuous speech recognition system =-=[1]-=- to generate phonetic segmentations. Based on these segmentations and probabilistic models we produce different pronunciation scores for individual or groups of sentences that can be used as predictor... |

31 | Combination of Machine Scores for Automatic Grading of Pronunciation Quality
- Franco, Digalakis
- 2000
(Show Context)
Citation Context ...bability distribution of the human and machine scores, it is necessary to resort to nonparametric methods. Useful nonparametric nonlinear methods to predict the human grades that we have investigated =-=[4]-=- are: neural networks, regression and classification trees and probability distribution estimation using scalar or vector quantization. 3.3. Nonparametric methods = ∑ Three possible implementations of... |

3 |
Modeling Intra-Word Pauses
- Franco, Neumeyer, et al.
- 1998
(Show Context)
Citation Context ...n problem, where we try to estimate, or predict, the value of the human grade by using a set of predictors. These predictors are the machine scores that we have presented in our previous work [2],[3],=-=[5]-=-. We investigate the use of MMSE estimation and classification methods to predict the human grade from a set of machine scores. We present alternative implementations of these methods based on nonpara... |

1 |
Classification and Regression Trees,” The Waxworks & Brooks/Cole Statistics/Probability Series
- Brahmin, Friedman, et al.
- 1984
(Show Context)
Citation Context ...puts will approximate the posterior probabilities Ph ( im) needed for classification in Eq. (3) [6]. Trees. Another approach to implementing the mappings is to use classification and regression trees =-=[7]-=-. In our case, a tree can be used to classify a vector of machine scores m to one of possible classes { t1, t2, …, tN} , each class representing a final node (a leaf) of the tree. The conditional dist... |