Phonetic labeling and stress assignment for turkish TTS

Abstract

Speech synthesis can be regarded as the automatic transformation of sentences from their text form into their speech waveform by machines. Synthesis systems can be divided into two major subsystems: (1) the natural language processing (NLP) subsystem where the conversion of input text into a linguistic representation, including phonetic and prosodic information, is done; (2) the speech signal processing subsystem where the generation of the output speech waveform using information provided by the NLP subsystem is handled (Ng, K., 1998; Semi, L., 1999). The NLP subsystem can be further divided into two parts: the transformation of text into its phonetic units which specify the sounds to be produced; the generation of prosodic parameters which specify how these sounds are to be produced (Ng, 1998; Semi, 1999). This study focuses on the conversion of text into its phonetic units and the extraction of stress information.