趨向補語在自然語音的使用研究

Directional Complements in Taiwan Mandarin Natural Speech

本文以台灣使用的國語語料探討趨向補語在自然對話的使用。口語語料先進行三種主要趨向補語語法結構的標記，再以詞頻的分析討論不同結構的使用狀況；結果與文本語料研究有相同也有相異之處。口語產製時語音表現可以反映結構特徵。因此，本文以統計方法測試有可能標示自然語音弱化的聲學參數。利用線性模型的殘差項進行的統計結果證實時長可以作為語音弱化穩定的量化指標。趨向補語保有原本空間表意的用法要比語法化的用法，在語音的聲韻標記上更為明顯。趨向補語的可能式，“得”與“不”的語音較相鄰的趨向補語為弱。同樣字詞用於趨向補語、動詞與形容詞，語法化的趨向補語用法明顯有語音弱化標記。

This paper is concerned with a specific kind of resultative verb complement construction in Mandarin Chinese, the directional complement construction. Instead of using texts or controlled laboratory data, we utilized a corpus of Taiwan Mandarin natural speech annotated with three types of directional constructions and investigated their usage in conversation. Some of the structures of directional complement constructions predicted by theories or observations were empirically confirmed in our frequency analysis, but some were not. In particular, we conducted statistical screening tests to explore relevant acoustic-prosodic features which may reflect the acoustic representation of directional components. The analysis was carried out by using residuals calculated by fitting a linear model of effects of speakers and syllables on the measurements of duration and intensity. The results suggest that duration is the most reliable acoustic-prosodic feature signifying phonetic reduction. Specifically speaking, the metaphorical use of directional constructions is prosodically less prominent than their original counterparts. Potentiality terms are more reduced than the neighbouring directional components. Phonetic reduction associated with syntactic prominence and semantic loss is also found in the use of directional components in contrast with the use of other grammatical categories such as verbs and prepositions. This paper not only gives a comprehensive overview of the authentic use of directional constructions in natural speech. It also provides a new piece of acoustic-prosodic evidence suggesting that the more grammaticalized the lexical items are, the more reduced they would appear in speech production.