語篇的基頻構組與語流韻律體現

An F0 Analysis of Discourse Construction and Global Information in Realized Narrative Prosody

本研究從語篇組織角度重新定義語篇韻律單位，依「階層式多短語語流韻律HPG 架構」(Hierarchical Prosodic Phrase Grouping) 之層疊與覆蓋關係，解構不同韻律格式語料，以音高的構組為例，證明口語語流的基頻聲學體現，乃由字調、韻律詞調、句法調、語篇聯繫位置、邊界延長、邊界停頓等成分系統性共構，並以各級韻律單位的相鄰與跨單位之對比與反差，證明較大韻律單位與語篇邊界效應亦為韻律語境之構組成分。各級韻律單位對韻律生成均有貢獻，層級貢獻度可透過統計分析釐清，不同韻律格式之體現反映於層級貢獻度的分布差異。研究方法則凸顯語料庫語言學及計算語言學解析動態超音段聲學參數相對性的重要與意義。

The aim of the present study is to show how the time varying F0 signal in paragraph prosody output is composed of contributions from multiple ranks of local between-unit concatenation and global higher-level layering. Discourse specified positioning, pre-boundary lengthening, and post-boundary pauses are also contributing variables. Both adjacency smoothing and higher level layering collectively make up the output F0, as supported by quantitative results. Results also reveal distinct patterns of phrase-level F0 height modulations within and between paragraphs, confirming the significance of phrase-level global prosody context in addition to tone-intonation interaction. Cross-speaker analysis of speech data varying in prosodic style show systematic patterns of contribution distribution by style, thus proving one base form is sufficient for both the planning and processing of surface variations. The study also shows why both methodology and interpretation must reflect the relative nature of suprasegmental cues, and how corpus analysis sheds new light on acoustic analysis.