Annotating Second Language Learner Speech: Reflecting on the Development of the J-AESOP Corpus

Abstract

This paper presents several proposals on annotating second language (L2) learner speech, reflecting on the author’s involvement in developing an L2 English speech corpus of Japanese speakers “J-AESOP”. The paper recommends broad (phonological) rather than narrow (phonetic) transcription of L2 speech because an accurate transcription of interlanguage phones is probably possible only after detailed acoustic analysis. The paper also discusses how existing automatic annotation tools can be applied to L2 speech and how L2-specifc speech phenomena (e.g., epenthesis, deletion, mispronunciation, disfluency) can be dealt with.