Stress is another important language-specific suprasegmental feature.
For Thai monosyllabic words, all content words are strongly stressed while all grammatical words are weakly stressed unless reinforced by emphasis. Thai polysyllabic words, however, have variable syllable-stress patterns such as (LL’), (LO’), (O’O’), and (O’L’) for bi-syllabic words, where L and O represent linking and non-linking syllables, and the symbol (’) is a stress marker given behind a stressed syllable. The linking syllable is a syllable having a vowel /a/ and a glottal stop. Thubthong et al. (2001a) investigated detecting stressed syllables in Thai polysyllabic words. Prosodic features, related to F0, duration, and energy, were calculated for different speech units including vowel, syllable, and rhyme units. An artificial neural network was used as the classifier. The model based on rhyme units achieved the best accuracy; an 84% stress detection rate. In Thai, there exist many homographs, which can often be distinguished only by context and/or stress. This is similar to some words in English, such as ‘‘present’’, which is pronounced with two different stress patterns, where each carries a different meaning. An analysis and modeling of stress in Thai phrases was carried out by Potisuk et al. (1996a). Acoustic parameters such as syllable rhyme-duration, F0, and energy were used to construct a Bayesian classifier, which achieved a 96% detection accuracy. The obtained prosodic constraints were
additionally incorporated in a syntactic parser based on a constraint dependency grammar (CDG) for Thai (Potisuk and Harper, 1996b).