Automatic phonetic transcription of dialectal variance in Catalan

Abstract

This project, which blended two fields of linguistics, language technology and dialectology, accomplished three main goals. Firstly, it improved TexAFon, a rule and knowledge based speech processing tool and upgraded it to handle eight northwestern Catalan varieties. Secondly, it enabled the tool for ASR use. Thirdly, it prepared TexAFon for use as an automatic pronunciation training system. The dialects were chosen as a result of a thorough research both in dialectology and language technology so as to define phenomena for Text-To-Speech (TTS) and Automatic Speech Recognition (ASR) modeling. The result of this project is an updated version of TexAFon, which can now be used for TTS and ASR in several Catalan dialects, and provide the basis for its use as an automatic pronunciation training/nsystem. Furthermore, this work also shows the upcoming need for updated dialectological studies in order to get present information about phenomena in different Catalan dialects.