Pause Insertion Based on a Morphosyntactic Parser for Brazilian Portuguese Text-to-Speech Systems

摘要

Based on a morphosyntactic parser a set of rules for inserting silence pauses in synthesized sentences coming from a Brazilian Portuguese text-to-speech (TTS) system is proposed. Such rules are determined as a function of the morphological class of words that compose the synthesized sentence, aiming to improve the TTS prosody. In order to assess the quality of the synthesized speech before and after inserting silence pauses, comparison category rating (CCR) listening tests are accomplished. Comparison mean opinion score (CMOS) results point out a considerable improvement in speech naturalness.