Development of a Mandarin monosyllable test material with homogenous items (I): Homogeneity selection

Abstract

Conclusion. A homogeneity selection could establish a homogenous resource for Chinese Mandarin monosyllabic test lists. Objective: This study aimed to develop a set of succinct and phonemically balanced monosyllabic Mandarin recognition test lists with good item homogeneity, in an effort to meet the need for Chinese Mandarin speech recognition test materials. Methods: In reference to the Chinese phonemes distribution probability, we developed the Phoneme Allocation Table, which decided the occurrence of all consonants, vowels, and semantic tones in each list. A total of 489 monosyllables were selected and organized into 30 lists, with 25 monosyllables in each. A two-factor design was used to screen homogenous items out. Sixty normal-hearing native Mandarin speakers participated in experiment. The performance-intensity (P-I) function curve of each test item was fit by logistic regression. Items were screened out as homogenous ones using the following criteria: 1) regression R value was not <0.9; 2) slope was between 2%/dB and 12%/dB; 3) threshold was between –8 dB HL and 10 dB HL. Results: The Phoneme Allocation Table was established as a framework of 30 lists, providing a basis on which phonemically balanced monosyllabic lists can be developed. In all, 342 monosyllables with good homogeneity were screened out and can be reorganized to establish monosyllabic test lists with good sensitivity.