Model-based automatic evaluation of second-language learner’s English segmental duration characteristics

Acoustical Science and Technology

Chatchawarn Hansakunbuntheung¹  Hiroaki Kato²  Yoshinori Sagisaka¹ 

[1] GITI/Language and Speech Science Research Laboratories, Waseda University;National Institute of Information and Communications Technology/ATR Media Information Science Laboratories

关键词: Segmental duration;    Automatic evaluation;    Second-language learning;    Linear regression model;   

DOI  :  10.1250/ast.31.267

学科分类：声学和超声波

来源: Acoustical Society of Japan

【 摘 要 】

References(33)In this paper, we propose a method of automatically measuring the segmental duration characteristics of a second-language learner’s speech as a means to evaluate language proficiency in terms of speech production. We propose the use of duration differences from native speakers’ speech as an objective evaluation score to evaluate the learner’s English segmental duration characteristics. To provide flexible evaluation without the need to collect any additional native-English reference speech, we employed predicted normalized segmental durations using a statistical duration model instead of measured raw durations of native reference speech. The proposed evaluation method was tested using English speech data uttered by multiple Thai-native learners’ groups with different amounts of experience of English study in English-as-an-official-language countries. An evaluation experiment showed that the proposed measure based on duration differences is strongly correlated with the amount of English study. Moreover, segmental duration differences revealed Thai learners’ speech-control characteristics such as stress assignment on word-final syllables. These results support the effectiveness of the proposed model-based objective evaluation.

【 授权许可】

Unknown