아나운서와 음성합성기의 뉴스 낭독 음성 비교 연구 -유사언어에서의 차이를 중심으로-

Abstract

I studied the paralanguage differences between a human announcer and TTS devices during the out-loud reading of news sentences, and arrived at the following results. First, the announcer’s intonation changed flexibly within a sentence whereas the intonation of the TTS devices was rather fixed at an either rising or falling tone. Second, the announcer stressed specific words within a sentence by changing her voice intensity, whereas the TTS devices maintained the same intensity throughout the sentence. Third, pause duration differed for both of the announcer and the “nVoice”-type TTS device but not for the “VoiceText”-type TTS device. In addition, the TTS devices made errors in reading by punctuating. Fourth, the announcer kept changing the pace within a sentence, whereas the TTS devices read all the words at an almost constant pace. Fifth, the announcer pronounced short and long vowels distinctively, but the TTS devices pronounced all vowels almost always using the same length. Sixth, the announcer widened her mouth when pronouncing open vowels, corresponding with the opening widths of given vowels, whereas the opening width of the TTS devices was smaller compared to that of the announcer’s. In sum, when reading aloud news sentences, the human announcer controlled the paralanguage features in a context-dependent manner, whereas the TTS devices tended to maintain the same paralanguage features throughout entire sentences.