Comparison of the prediction accuracy of machine learning algorithms in crosslinguistic vowel classification

Abstract

Machine learning algorithms can be used for the prediction of nonnative sound classification based on crosslinguistic acoustic similarity. To date, very few linguistic studies have compared the classification accuracy of different algorithms. This study aims to assess how well machines align with human speech perception by assessing the ability of three machine learning algorithms, namely, linear discriminant analysis (LDA), decision tree (C5.0), and neural network (NNET), to predict the classification of second language (L2) sounds in terms of first language (L1) categories. The models were trained using the first three formants and duration of L1 vowels and fed with the same acoustic features of L2 vowels. To validate their accuracy, adult L2 speakers completed a perceptual classification task. The results indicated that NNET predicted with success the classification of all L2 vowels with the highest proportion in terms of L1 categories, while LDA and C5.0 missed only one vowel each. Furthermore, NNET exhibited superior accuracy in predicting the full range of above chance responses, followed closely by LDA. C5.0 did not meet the anticipated performance levels. The findings can hold significant implications for advancing both the theoretical and practical frameworks of speech acquisition.

Acoustic Similarity Predicts Vowel Phoneme Detection in an Unfamiliar Regional Accent: Evidence from Monolinguals, Bilinguals and Second-Language Learners

“…Recall that cross-language acoustic similarity has been studied mainly to identify potential challenges in perceiving vowels by non-native and occasionally L2 listeners (Georgiou 2023). This endeavor has generally been motivated by the notion that acoustic classification patterns approximate patterns of perceived phonetic similarity.…”

“…Acoustic similarity between the realizations of phonemes across native and non-native languages has been used to predict cross-language perceptual categorization, which in turn may reflect challenges in the perception of speech in the target non-native language (for a recent review, see Georgiou 2023). In the case of vowels, a line of studies has found that cross-language acoustic similarity predicts performance in speech perception tasks by individuals inexperienced in the target non-native language, as well as by L2 learners with some experience (e.g., Gilichinskaya and Strange 2010;Escudero and Vasiliev 2011;Elvin et al 2014;Strange et al 2011;Georgiou 2023;Williams and Escudero 2014b). In a now established methodology, acoustic measurements, e.g., duration and first, second and third formant (F1, F2, F3) frequency values from tokens of native vowel categories act as input into a statistical classification model.…”

Acoustic Characteristics of Greek Vowels Produced by Adult Heritage Speakers of Albanian

“…The target words spoken by the speakers were extracted and processed in Praat [61] (for similar descriptions, see [62,63]). Spectrograms and waveforms were visually examined to identify key acoustic features, enabling the measurement of vowel boundaries, including formant frequencies and vocalic duration.…”

Marketing Strategy of Private Enterprises Based on Bayesian Dynamic Panel Model of Machine Learning Algorithms

“…At the same time, in the study of practical problems, researchers often pay more attention to both ends of the conditional distribution of dependent variables than the average impact of independent variables on dependent variables. Georgiou (2023) pointed out that high concentrations of pollution have a more serious impact on human beings and ecosystems. Therefore, it is essential to study this group of countries to formulate effective environmental protection policies.…”