네이버 파파고와 구글 번역기 음성 산출의 연음 현상에 관한 연구

Abstract

This study investigated the pronunciation of liaison of TTS provided by Naver Papago and Google Translator. We looked at the pronunciation of the liaison of the TTS provided by the two translators, and suggested a way to provide a more natural pronunciation of the TTS from a Korean linguistical perspective. First of all, through research, we found that the TTS provided by Naver Papago and Google Translator is generally faithful to the standard pronunciation, but there is some mixed reality pronunciation. However, it is pointed out that the standard pronunciation and the real pronunciation are implemented non-systematically. In order to improve this situation, this study proposed standard pronunciation and real pronunciation information for each unit of verse, suggesting a more natural approach to TTS. In addition, since the acceptance degree of the standard pronunciation and the actual pronunciation of the speaker may be different, it is suggested to design the TTS system considering the actual pronunciation acceptance degree.

This study investigated the pronunciation of liaison of TTS provided by Naver Papago and Google Translator. We looked at the pronunciation of the liaison of the TTS provided by the two translators, and suggested a way to provide a more natural pronunciation of the TTS from a Korean linguistical perspective. First of all, through research, we found that the TTS provided by Naver Papago and Google Translator is generally faithful to the standard pronunciation, but there is some mixed reality pronunciation. However, it is pointed out that the standard pronunciation and the real pronunciation are implemented non-systematically. In order to improve this situation, this study proposed standard pronunciation and real pronunciation information for each unit of verse, suggesting a more natural approach to TTS. In addition, since the acceptance degree of the standard pronunciation and the actual pronunciation of the speaker may be different, it is suggested to design the TTS system considering the actual pronunciation acceptance degree.