A Rule Based Syllabification Algorithm for Sinhala

R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 438 – 449, 2005. © Springer-Verlag Berlin Heidelberg 2005 A Rule Based Syllabification Algorithm for Sinhala Ruvan Weerasinghe, Asanka Wasala, and Kumudu Gamage Language Technology Research Laboratory, University of Colombo School of Computing, 35, Reid Avenue, Colombo 7, Sri Lanka arw@ucsc.cmb.ac.lk, {awasala, kgamage}@webmail.cmb.ac.lk Abstract. This paper presents a study of Sinhala syllable structure and an algo- rithm for identifying syllables in Sinhala words. After a thorough study of the Syllable structure and linguistic rules for syllabification of Sinhala words and a survey of the relevant literature, a set of rules was identified and implemented as a simple, easy-to-implement algorithm. The algorithm was tested using 30,000 distinct words obtained from a corpus and compared with the same words manually syllabified. The algorithm performs with 99.95 % accuracy. 1 Introduction Syllabification algorithms are mainly used in text-to-speech (TTS) systems in produc- ing natural sounding speech, and in speech recognizers in detecting out-of-vocabulary words. The key objectives of this study are to identify the syllable structures in modern Sinhala language and to define an algorithm to syllabify a given Sinhala word to be used in our TTS system. Syllabification algorithms have been proposed for different languages including English, German, Spanish and Hindi, among others. Although a few researchers have documented attempts at syllabifying modern Sinhala words in the Linguistics literature, this is the first known documented algorithm for Sinhala syllabi- fication and certainly the first evaluation of any syllabification scheme for Sinhala. Languages differ considerably in the syllable structures that they permit. For most languages, syllabification can be achieved by writing a set of declarative grammatical rules which explain the location of syllable boundaries of words step-by-step. It has been identified that most of these rules adhere to well known theories such as the Maximum Onset Principle and the Sonority Profile. The association of consonants with the syllable nucleus is derived by the Maximum Onset Principle (MOP). Maximum Onset Principal: First make the onset as long as it legitimately can be; then form a legitimate coda [2]. Sonority Profile: The sonority of a syllable increases from the beginning of the sylla- ble onward, and decreases from the beginning of the peak onwards [2]. Sonority is related to the acoustic intensity of a sound. Thus, by measuring the acoustic intensities of sounds, the sonority of a sound can be estimated [1]. The classes of vowel and consonant sounds (segments) that are usually distinguished along this dimension are listed in the order of increasing sonority, and this list is re- ferred to as the Sonority Scale [2]. A Rule Based Syllabification Algorithm for Sinhala 439 Sonority Scale: Obstruents – Nasals – Liquids ([l, r] etc.) – Glides ([w, j]) etc.) – Vowels Many syllabification algorithms have been developed based on these two theories. For example, the Festival Speech Synthesis System (the framework we use in our Sinhala TTS) by default syllabifies words by finding the minimum sonorant position between vowels [3]. Another sonority scale based …

Content truncated to 500 words