Sri Lankan country component of PAN Localization is the only component looking at developing a local language text to speech synthesis system. University of Colombo School of Computing, the PAN Localization collaboration institution, already has a very long and rich  

tradition of localization. this work has been going on for more than a decade...

Recently, with localization and ICT proliferation interest sparked through ICT Agency of Sri Lanka, more work has been evolved.  This complements the PAN Localization project as well. Current work, through this project, focuses on advanced localization applications.  UCSC is already been doing work in speech recognition with cooperation of JICA.  Within PAN Localization, the group will be focusing on developing Sinhalese text corpus, tri-lingual Sinhalese-English-Tamil dictionary, Optical Character Recognition System and Text-to-Speech system (in addition to developing some basic standards, e.g. for Sinhalese collation)

Even though UCSC has considerable expertise in Computational Linguistics and text processing, it has only limited experience in text to speech synthesis.  There have been work in concatenate synthesis and also work on Sinhalese phonetics and phonology but most of this work has not been put together. There has also been a limited exposure to acoustic phonetics, as a bridging

science between phonetics and phonology in Linguistics, on one end, and computer speech synthesis on the other.

Sri Lankan team of PAN Localization had identified text to speech system as a need of the country to effectively enable ICTs for its population.  Thus the component holds an essential focus for this team.  With missing expertise, it was therefore decided to hold a training on “Phonetics and Phonology for TTS”. 

Objectives of the training were to give requisite  background to the Sri Lankan PAN Localization  team and others interested on Phonetics,  Phonology,  Acoustic Phonetics and how all  this  fits in with TTS process. The team, predominantly  contains computer science graduates, including a  junior and a very senior linguist. Thus, along with  the basic concepts of physical speech processes

and the underlying rules, they had to be taught detailed acoustic phonetics since it is essential for text to speech systems. Also, the knowledge of relevant computer tools to analyze the speech signal was also required.

The books used as the main sources of knowledge for training purposes were:

1.   Clark & Yallop, "Phonetics and Phonology"

2.   Ladefoged, "A course in Phonetics"

3.   Pickett, "Acoustics of Speech Communication"

4.   Goldsmith, "Auto-segmental and Metrical Phonology"

5.   Dutoit, "An Introduction to Text to Speech Synthesis" (Dutoit)

In this training program, 40 hours of course work was designed to cover the following very useful study areas:

	Phonetics
		Introduction to anatomy and speech of production
		Articulatory phonetics, sound classification and transcription
			Air stream mechanics
			Voicing
			Articulation: manner, place
			IPA and Sinhalese writing system
		Supra-segments
			Stress
			Intonation
		Acoustic Phonetics
			Source-filter theory of speech production
			Acoustic cues of vowels
			Acoustic cues of consonants
			Co-articulation
			Supra-segment effects
	Phonology
		Introduction to auto-segmental phonology
		Phonological rules and ordering
		Syllabification
		Stress and metrical phonology
	Text to Speech Synthesis
		Architecture of a synthesizer
		Synthesis Techniques
		Diphone Synthesis