台灣華語語意預測度之韻律標記

Abstract

本研究探討台灣華語語意預測度高低是否具有不同的聲學呈現。文獻指出，語意預測度較高的語境中，語音單位時長較短，音強較低，而且發音時，母音位於以第一、第二共振峰架構空間圖較為中間的位置。本研究聚焦聲調語言，除了檢驗文獻有關元音的發現，更探討語意預測度是否影響基頻的聲學表現。實驗前測招募十二位受試者，工作內容包括猜測句中遭刪除的標的詞，並根據句子與標的詞間的語意預測度評分。答對率及評分結果顯示，本研究採用的語意對比句確實可信。構音實驗招募二十位發音人，有一問句引導發音人唸出標的句。實驗結果與文獻研究一致之處在於，語意預測度較高時，元音時長較短，音強較弱，在第一、第二共振峰架構空間圖中，元音較靠近中央位置。本研究更進一步指出，基頻範圍也受語意預測度影響。語意預測度較高時，基頻範圍縮小。此外，語意預測度較低時，性別反映顯著的基頻範圍差異；但預測度較高時，此差異卻被弭平。總而言之，本論文為語意與語音之聯結提供了有力的證據。人類語言是一套複雜的系統，說話者不光傳遞一連串的語音訊號，更對文字以外的資訊進行編碼，以有效傳達更多訊息。

This thesis investigated the effect of semantic predictability on acoustic realizations in Taiwan Mandarin. Most previous studies have been guided by the assumption that in high predictability conditions, speech units tend to be reduced in the temporal and spectral qualities. This thesis focuses on a tone language in an attempt to not only examine previous findings about vowels but also look into possible effect of semantic predictability on F0. Prior to the recording, twelve subjects were asked to guess the missing target word in a fragment and to rate the predictability of the sentence. The results of correctness percentage and predictability rating both confirmed the reliability of the present corpus. A set of pretested sentences were elicited from twenty native speakers of Taiwan Mandarin, inclusive of ten males and ten females. The results shown that, consistent with previous findings, words were shorter in vowel duration, lower in the intensity, and more centralized within the vowel space when they were embedded in high predictability contexts. Male and female speakers both performed the same reduction process systematically. F0 excursion was also found strongly affected by semantic predictability; specifically, it was much smaller in high predictability conditions. Moreover, gender difference in the extent of F0 excursion was mediated in high predictability conditions. This thesis provides additional evidence for the acoustic correlates of semantic predictability in speech production and sheds light on the prosodic encoding of semantic structures in tone languages.