泰國學齡前兒童聲調之初步研究

Interlanguage Tone Patterns in Thai Pre-school Children: A Preliminary Corpus Analysis

本文主要是對於泰國曼谷中華國際學校11名幼童學習華語提出初步的聲調中介語研究。本案所有學習語料庫建構於Praat聲學聲檔中，語料庫自動分成17層，其中涵蓋授課老師和參與幼童的口語語音，並且採用機器學習方式半自動處理聲學語料（人工處理僅佔30％，系統自動標註語言學標記佔70％）。本研究將幼兒詞彙中的音節數量（單音節、雙音節及多音節）搭配聲學基頻（F0）做一初步觀察。目前結果顯示為泰國幼童本身母語為五調聲調多音節語言，比華語複雜，因此他們口語產製華語聲調並沒有任何困難，而且他們對於雙音節詞彙有相當高的使用比例。本文建議應以中介語作為學習外語的過渡中介期，主因在於這群泰國幼童產生的聲調偏誤類別，聲學基頻參數顯示並非來自於泰語母語聲調，而是偏向近似華語的聲調調型。由此可見，泰國幼童或許刻意建構一個獨立聲調新系統，作為學習新語言的中介過渡期。未來研究將會在華語聲調辨識上執行另一項語音辨識實驗，亦即將所有聲檔以低頻波過濾所有的語意詞彙，以求更精準的譯寫正確率。

The aim of this paper is to present a preliminary analysis by providing a detailed corpus study of interlanguage tone patterns made by 11 pre-school children in Thailand who learn Mandarin as their global language at Thai-Chinese International School in Bangkok, Thailand. The novelty of this work is to collect and analyze some data from a highly reliable corpus, and provide Praat acoustic parameters for the tone distribution by looking at various tone combinations measured by plotting the F0 variations in Mandarin. These units involved in the errors are classified by segmenting word shapes into monosyllabic, disyllabic or multi-syllabic. Evidence from the data shows that Thai children have no problem producing Mandarin tones, and are more likely to utter disyllabic lexicons. Different from the traditional analysis on the error patterns, which argues that all the error patterns were made based on the learners' mother tongue, this study has suggested that interlanguage can be viewed as the transitional process between the mother tongue and the target language, and those tone error patterns are not fully related to their mother tongue, possibly suggesting that Thai children might have constructed an independent linguistic system gradually towards Mandarin. In the near future, a phonetically perceptual task in a low-pass filter for reserving tone information will be required in the hope of eliminating the transcribers' bias to rely on the lexical information.