ISCA Archive Interspeech 2023
ISCA Archive Interspeech 2023

A Unified Framework to Improve Learners' Skills of Perception and Production Based on Speech Shadowing and Overlapping

Nobuaki Minematsu, Noriko Nakanishi, Yingxiang Gao, Haitong Sun

A unified framework to improve learners' skills in perceiving and producing L2 sounds is demonstrated based on speech shadowing and overlapping. Speech shadowing is a training method, where learners are asked to reproduce a given model speech (M) as immediately as possible, and it was proved to be effective in enhancing their L2 speech perception. After several trials of shadowing, the learners are provided with M's script to continue shadowing with no delay, called overlapping. By comparing the shadowing speech (S) and the script-shadowing speech (SS), shadowing breakdowns are measured sequentially, which can characterize listening breakdowns. By comparing M and SS, the prosodic and segmental gaps are analyzed sequentially and presented visually to learners along with imitation scores. All the tasks are implemented as interactive speech games, which help learners to become more proficient in L2 speech perception and production.