台語口語及書面語體的多面向分析

A Multi-dimensional Analysis of Spoken and Written Taiwanese Register

這篇論文是以語料庫為本的社會語言學研究。本文嘗試辨認影響台語語體變異的深層言談面向，同時刻畫語體的篇章關係。本研究採用Biber 的「多面向方法」。本研究建立的語料庫包含口語語體（9類，94篇）及書面語體（8類，91篇），總計144,942詞。經過相關的統計分析後，總共得出五個深層言談面向：人際關係vs. 表達資訊；個人情感的表達；說服：邏輯vs. 空間的連接；敘述；具親和力的說明vs. 詳細報導。本文利用語體的面向分數來刻劃語體的篇章關係。研究結果顯示語體的變異需要透過多面向的分析來掌握，因為任何單一面向都不能絕對的區分口語及書面語體的差異。

This paper is a corpus-based study of sociolinguistics. The aim of this study is twofold: (1) to identify underlying dimensions of register variation in Taiwanese, and (2) to characterize the textual relations among different registers. This study adopts Biber’s (1988) multi-dimensional approach to register variation. The corpus comprises nine spoken registers (94 texts) and eight written registers (91 texts), totaling 144,942 words. Texts are segmented, tagged, and counted by computational tools. The distribution of 36 linguistic features in these texts is subjected to factor analysis. Five underlying dimensions are identified in this study: interpersonal vs. informational; personal expression of emotion; persuasion: logical vs. temporal linking; narrative; and involved exposition vs. precise reportage. Textual relations among these registers support the interpretation of dimensions. Since no single dimension is capable of explaining all of the register variation in Taiwanese, the results of this study show that register variation are best captured by multi-dimensional construct.