In a polyglot speech synthesis, it is possible to use one language resource for another language. However, if the adaptation is not implemented carefully, the foreignness of the sound will be too noticeable for the listeners. This paper presents the analysis of respondents’ acceptance of a series of listening tests. The research goal was to find out in the absence of phonemes of a particular language, would it be possible for the phonemes to be replaced with another language’s phonemes. This will be especially beneficial for under-resourced language either in the case for 1) the language has not yet well researched into or 2) the language has not well documented in the required media. Preliminary studies were conducted to construct phoneme confusion matrices. The confusion study was observed based on the consonants’ position in syllable structure: onset and coda. These studies were then compared to similar studies to find possible overlap among them. Then, based on the outcome, two perceptual tests have been conducted to observe the applicability of phoneme substitutions. The first test was to observe the effect of phonemes substitution during the intelligibility test for individual words. The second test was to evaluate whether context influenced perception based on whether respondents noticed phoneme substitution on a word in a series of words. From these experiments, it can be concluded that it is possible to do phoneme substitution but with a certain condition. From significance testing, it was found that phoneme substitution may not be suitable to be implemented for onset position but can be applied for coda position provided the context is available.
Bibliographical noteFunding Information:
Funding Statement: This manuscript publication fee is partially funded by the Universiti Sains Malaysia’s Short term grant no: 304/PKOMP/6315273.
© 2022, Tech Science Press. All rights reserved.
- Language similarity
- Natural language processing
- Phoneme analysis
- Polyglot under-resourced language
- Speech analytics
ASJC Scopus subject areas
- Theoretical Computer Science
- Computational Theory and Mathematics
- Artificial Intelligence