Impacts of multicollinearity on CAPT modalities: An heterogeneous machine learning framework for computer-assisted French phoneme pronunciation training

Abstract

Phoneme pronunciations are usually considered as basic skills for learning a foreign language. Practicing the pronunciations in a computer-assisted way is helpful in a self-directed or long-distance learning environment. Recent researches indicate that machine learning is a promising method to build high-performance computer-assisted pronunciation training modalities. Many data-driven classifying models, such as support vector machines, back-propagation networks, deep neural networks and convolutional neural networks, are increasingly widely used for it. Yet, the acoustic waveforms of phoneme are essentially modulated from the base vibrations of vocal cords, and this fact somehow makes the predictors collinear, distorting the classifying models. A commonly-used solution to address this issue is to suppressing the collinearity of predictors via partial least square regressing algorithm. It allows to obtain high-quality predictor weighting results via predictor relationship analysis. However, as a linear regressor, the classifiers of this type possess very simple topology structures, constraining the universality of the regressors. For this issue, this paper presents an heterogeneous phoneme recognition framework which can further benefit the phoneme pronunciation diagnostic tasks by combining the partial least square with support vector machines. A French phoneme data set containing 4830 samples is established for the evaluation experiments. The experiments of this paper demonstrates that the new method improves the accuracy performance of the phoneme classifiers by 0.21 − 8.47% comparing to state-of-the-arts with different data training data density.

A RTL Implementation of Heterogeneous Machine Learning Network for French Computer Assisted Pronunciation Training

“…where ξ i and ξ i are the slack variables. µ i ≥ 0, µ i ≥ 0, α i ≥ 0 and α i ≥ 0, which correspond to the columns of µ µ µ, µ µ µ , α α α and α α α , are the lagrange multipliers and can be solved by building the dual problem of ( 8) with the Karush-Kuhn-Tucher constraints [27]. The desired coefficient matrix W W W (2) of the second layer are obtained by computing the partial derivatives of ( 9) with respects to W W W (2) , b, ξ i and…”

“…The decision of the system is made by comparing the output of the detector y, which is the diagnosis score corresponding to the utterance quality, with a threshold η to feedback the diagnosis result. This work trains the detectors through a heterogeneous process presented in [27]. It consists of partial least square (PLS) regression and soft-margin support vector machines.…”

“…Concerning the pronunciation diagnosis algorithm, a high-accuracy and low-consumption classifier is desired to balance accuracy and efficient performance. A recently-proposed heterogeneous machine learning CAPT framework [27] is therefore selected. For the reason that the phoneme utterances are made from the base vibrations of vocal cords through resonance chambers (buccal, nasal, and pharyngeal cavities) [28,29], the predictors of the phoneme feature vectors are highly probably collinear, resulting in a multicollinearity problem.…”

“…Although it is usually difficult to figure out a precise mathematical model to explain the fundamentals ofal least squar a certain pattern recognition problem, research indicates that suppressing the multicollinearity by using some suitable method is helpful to improve the pattern discriminability [30][31][32]. Yanjing et al [27] estimate the condition indices of a French phoneme utterance spectrum set, and 87.27% of its elements exceed 10. This means that the predictor dependencies start to affect the regression estimates [33].…”

Correction: Impacts of multicollinearity on CAPT modalities: An heterogeneous machine learning framework for computer-assisted French phoneme pronunciation training

“…

After publication of this article [1], concerns were raised that information regarding the ethical approval and participant consent information for the construction of the phoneme dataset was omitted from the article. The correct information is: The construction of the phoneme dataset was approved by the School of Foreign Studies at the Capital University of Economics and Business (Beijing, China).

…”