Second Language Studies: Acquisition, Learning, Education and Technology

Tokyo, Japan
September 22-24, 2010

Speech Analysis for Automatic Evaluation of Shadowing

Dean Luo (1), Yutaka Yamauchi (2), Nobuaki Minematsu (1)

(1) The University of Tokyo, Tokyo, Japan
(2) Tokyo International University, Saitama, Japan

This paper presents acoustic analysis for the purpose of automatic evaluation of shadowing speech. We use selfchecked scores of understanding, manual prosodic scores, and TOEIC scores as reference scores of learners’ shadowing speech, and compare these scores with automatic scores based on acoustic features that can reflect phoneme intelligibility and prosodic fluency in terms of intonation, and rhythm. We also examine the differences of personal-best shadowing, shadowing after the transcription is shown and reading-aloud of the same contents. Experimental results show that learners’ understanding of contents in shadowing affects segmental intelligibility and prosodic fluency of their shadowing productions. A multiple regression model that combines different features can better reflect learners’ understanding of the contents of shadowing and other reference scores, and thus suitable for automatic evaluation of shadowing.

Full Paper

Bibliographic reference.  Luo, Dean / Yamauchi, Yutaka / Minematsu, Nobuaki (2010): "Speech analysis for automatic evaluation of shadowing", In L2WS-2010, paper P2-8.