自然對話的處理單位：法語與漢語的比較研究

Processing Units in Conversation: A Comparative Study of French and Mandarin Data

人類口語產製主要目的是為了傳達訊息給予聽話者，進而達到溝通目的。因此，長段話語切分為較小的處理單位將更易於建構意義清楚的言談架構。在自然對話中，此類型的處理單位可能參照語意、句法或韻律層次的結構。過去文獻曾提出不同的理論嘗試解釋言談切分的運作，但多半屬於質化分析研究。本研究利用以語料庫為本的量化資料，根據語意訊息結構（言談單位）、韻律表現（音韻單位）和詞類標記序列（詞組）標記個別處理單位之邊界。以此為根基比較台灣華語和法語在這三種處理單位上的結構表現，並檢視這三種處理單位之間的互動關係。本研究以語料庫語言學方法為兩個類型迥異的語言建築一個對等的比較平台，進而針對自然對話的語言表現，作跨語言之量化比較研究。

Human spoken language production is directed towards communication delivering comprehensible information to recipients. Speech segmentation into small units efficiently enhances a sensible and interpretable discourse structure. Such processing units in real-life communication may be applied to semantic, syntactic, or prosodic structures. Previous studies have proposed various theories of speech segmentation, mainly based on qualitative analyses. The present study utilizes corpus-based quantitative data to examine how conversational speech in French and Mandarin is structured in terms of three different processing units, and how these units interact with one another. Unit completion location was identified by semantic structure (discourse unit), prosodic pattern (prosodic unit), and sequences of parts of speech (chunk). Quantitative analyses for both languages were carried out by applying comparable processing procedures. This article presents our efforts to establish a dataset for two typologically diverse languages, and to carry out quantitative comparative studies of processing units in face-to-face conversation.