A Finite State Machine Model to Determine Syllables of Indonesian Text

Abstract

Syllable segmentation is an important process in the development of realistic visual speech synthesis for the Indonesian language. The process is to determine the syllables in a text by applying certain rules. In this study, we develop a rule-based syllable segmentation to convert Indonesian text into syllables. The rules are represented in the form of a finite state machine (FSM). The FSM model depends on the combination of vowels and consonants. The syllable peak indicated by the sound of a vowel, while the consonant sound represents a syllable valley. In a syllable, there is only a syllable peak, but the syllable valley can be more than one. Therefore, the number of syllables can be determined by looking at the vowels present in a word. The rules can be formulated by looking at the syllable pattern. The pattern can only be a vowel sound, a vowel with a consonant sound, and a vowel with two consonant sounds and others. The evaluation is done by entering Indonesian texts to the FSM model. The result shows that the FSM model could determine the syllables properly.