語音辨識系統上的模仿、比較和回饋方法對華語學習者「發音」的影響

對於華語學習者而言，溝通是其學習中文的重要目的，而要能清楚傳遞訊息，進行溝通，勢必需要正確的發音。伴隨著網路科技及語音辨識系統的發展，電腦輔助發音學習正在流行，由傳統教室透過老師大班教學的方式轉變成使用新穎的資訊科技與學習理論的結合教學，讓學生可以自主學習，老師在課堂上可以將教學重心放在更高層次的教學。本研究針對學生的學習，提出使用聽後模仿、比較辨別與模仿回饋方法，結合語音辨識技術，進行電腦輔助發音教學，幫助學習者口說學習。研究方法間的結合，是否能協助學習者，在進行口說單字練習時，可以快速矯正其錯誤發音，增強學習者自主學習的成效。本研究的實際研究對象為60位外國學習者。透過隨機分派的方式，分成四組，分別針對不同方法的結合，進行比較，並了解學習者在學習後是否有所進步。研究結果顯示：一、參與者的背景可能影響口說學習成效，在本研究的學習者，來自漢字學習圈的平均成績高於非漢字學習圈的學習者。二、集合不同方法融入於電腦輔助口說學習的方式，讓各個組別在後測成績上均有進步。三、將三種方法結合使用的實驗組一，在口說成效測驗上顯著高於只有使用聽後模仿方法的控制組。四、在滿意度問卷的調查上，學習者大多給予肯定的評價。控制組的分數，略低於其他組別。最後，研究者針對研究結果與訪談內容進行討論，並提出幾點建議與未來研究方向以供日後研究與參考。

關鍵字

語音辨識；聽後模仿；電腦輔助發音學習；比較辨別；模仿回饋

並列摘要

For all the Chinese learners, the attainment of successful communication is essential. And the correct articulation/pronunciation is then necessary if we want to get our meaning across in a lucid style. Along with the advancement of technology and the development of speech recognition device, Computer Assisted Pronunciation Teaching (CAPT) is now under way and incorporated into teaching which conjoins both information technology and pedagogical theory. This sort of technical invention as a medium of pronunciation diagnosis not only helps students benefit from the pleasure of independent learning but also allows teachers to focus more on the quality of the teaching materials per se. This study proposes three strategies as what follows: Speech Imitation, Speech Discrimination, and Automatic Feedback after speakers imitate the correct sound given. Besides, Automatic Speech Recognition (ASR) would be applied in conjunction with these strategies to facilitate CAPT and also encourage learners to practise Chinese pronunciation. This research aims to argue that the different combinations of strategies may shed light on the correct sound acquisition of learners. Whilst practising monosyllables’ pronunciation, the deployment of the strategies are found to be helpful in rectifying their mispronunciation to several levels. The participants in this research include 60 foreign learners of Chinese language, and they had been randomised into 4 groups. Each group is designed to undergo a specific combination of strategies. Lastly, we’d evaluate and compare the learning effect of speakers amongst the four groups. The result shows that: 1. The background of participants would influence the Chinese pronunciation acquisition. Those who are of the Sino-sphere (Chinese cultural sphere) usually have a more desirable performance than those who aren’t. 2. Each group makes progress in the posttest, yet the degree of progress varies across groups. 3. The upshot of the pronunciation test of the group which taps into all three strategies is more satisfactory than that of the group which uses the sole strategy ‘Automatic Feddback.’ 4. From the questionnaire in Appendix, we know that most participants have positive feedback towards this system. In the last chapter, we will further our discussion based on the conclusion and the interviews with participants, and summarise this paper with some advices and potential research directions..