華語學習者語料庫語音偏誤研究：以不送氣舌尖後翹舌音、舌根鼻音和輕聲為例

A study of phonetic errors in Chinese learner corpus: Non-aspirated retroflex affricate, velar nasal, and neutral tone

本研究旨在分析各潛在因素與華語學習者語音偏誤之相關性。華語的「聲、韻、調」別具特色，形成這個語言獨特之美，但往往也是華語學習者學習華語時的一大難點。近三十年來，華語學習語音表現和教材教法，一直是學者研究重點。然而，先前研究或比較語言音韻系統異同，或以少數學習者為調查對象，缺乏偏誤發生頻率和各變項相關性等實證數據可做為教學應用參考。因此，本研究運用國立臺灣師範大學「華語為第二語口語語料庫」語音資料，進行調查分析。研究結果發現：（1）不送氣舌尖後翹舌音偏誤與學習者母語背景有關，偏誤率由高而低分別是日語母語者、英語母語者、韓語母語者。（2）舌根鼻音偏誤和中文等級程度有關，進階程度學習者偏誤率高於基礎程度學習者。亦與詞彙位置有關，詞首位置偏誤率高於詞尾位置。（3）輕聲偏誤與學習者母語背景有關聯，偏誤率由高而低分別是日語母語者、韓語母語者、英語母語者。亦與性別有關，女性學習者偏誤率高於男性。

This study aims to analyze the correlation between various potential factors and phonetic errors among L2 Chinese language learners (CLLs). Mandarin's distinctive phonetic features, including initials, finals, and tones, contribute to the language’s unique beauty but often pose challenges for CLLs. Over the past three decades, scholars have consistently focused on the CLLs’phonetic performance and instructional methods. However, the previous studies either compare phonological systems across languages or involve a limited number of learners as subjects, lacking empirical data on error occurrence frequency and the correlation with various variables that can be used as references for teaching applications. To fill this gap, this study utilizes phonetic data from the National Taiwan Normal University's“Spoken L2 Chinese Learner Corpus”for investigation and analysis. Findings reveal: (1) The occurrence of non-aspirated retroflex affricate error correlated with learners’native language backgrounds, with higher error rates observed among Japanese CLLs, followed by English CLLs and Korean CLLs. (2) Errors in velar nasals are associated with learner’s Chinese proficiency level, with advanced learners exhibiting higher error rates compared to beginners. Additionally, error rates are influenced by word position, with higher rates observed at word beginnings than endings. (3) Errors in neutral tones are related to learners' native language backgrounds, with higher error rates observed among Japanese CLLs, followed by Korean CLLs and English CLLs. Gender also plays a role, with female learners exhibiting higher error rates than males.