Multi-dialect phonetisation for Irish text-to-speech synthesis : a modular approach

Abstract

This thesis describes the development of a modular multi-dialect phonetiser for the Irish language. Phonetisation (or grapheme-to-phoneme conversion), the process in which ordinary text is converted into a transcription of how that text might be pronounced, is a key step in the process of text-to-speech (TTS) synthesis, the conversion of written text to artificial speech. Phonetisation for Irish poses a number of challenges, including a complex (but transparent) orthographic system, the absence of a standard dialect, and the lack of prerequisite technological resources for the language. For example, phonetisation typically uses a combination of letter-to-sound (LTS) rules and a pronunciation lexicon or dictionary, but no dialect-specific pronunciation lexicon is available for Irish.