Language Model-Based Emotion Prediction Methods for Emotional Speech Synthesis Systems

Authors: Hyunwook Yoon, Ohsung Kwon, Hoyeon Lee, Ryuichi Yamamoto, Eunwoo Song, Jae-Min Kim, and Min-Jae Hwang

Section 4.3.2 Subjective listening tests

Section 4.3.3 Paragraph-level emotional speech synthesis

  • Audio Samples
  • System S3 S4single S4multi

    "제비는 혼잣말을 했다." | Neutral
    "Swallow talked himself." | Neutral

    "그녀는 통 말을 안 해." | Anger
    "She doesn't talk to much." | Anger

    "게다가 바람둥이가 아닐까 싶어." | Anger
    "Besides, I think she's a playgirl." | Anger

    "확실히 바람이 불 때마다 갈대는 몸을 숙이며 아주 우아하게 인사했다." | Neutral
    "Certainly, whenever the wind blew, the reed bowed and greeted very gracefully." | Neutral

    "제비는 계속 말했다." | Neutral
    "The swallow kept talking." | Neutral

    "그녀는 너무 집에만 붙어 있어." | Anger
    "She's too attached to the house." | Anger

    System S1 S4single S4multi

    "옷매무새를 고치며 남자가 달갑지 않다는 투로 물었다." | Neutral
    "The man asked in an unwelcome tone while fixing his clothes." | Neutral

    "무슨 일이시오?" | Anger
    "What's going on?" | Anger

    "그리고 훈계조로 한마디 덧붙였다." | Neutral
    "And he added the words with admonishing tone." | Neutral

    "방에, 들어오기 전에, 노크를 하셔야죠!" | Anger
    "Before you come into the room, knock on the door!" | Anger

    "그의 등뒤 어둑한 방안에서 하얗고 큰 새 같은 것이 부산스럽게 움직이는가 싶더니 낭랑하고 명랑한 목소리가 들려왔다." | Neutral
    "In the dark room behind his back, I thought something like a white and big bird was moving busily, and a cheerful voice came out." | Neutral

    "특히 부부가 사는 집에 들어올 때는 말이에요..." | Joy
    "Especially when you come into a married couple's house..." | Joy


    Work performed with nVoice team, Clova Voice, Naver Corp.