ISCA Archive SLaTE 2017
ISCA Archive SLaTE 2017

A pipeline for automatic assessment of foreign language pronunciation

Aku Rouhe, Reima Karhila, Heini Kallio, Mikko Kurimo

We illustrate our prototype pipeline for automatic analysis and assessment of foreign language speech. It uses automatic speech recognition as a preprocessing step for phonetic analysis and predicts the grade that human experts would give for the utterance. The work is applied in two educational research projects: DigiTala for L2 Swedish for Finnish-speaking high school students, and SIAK ”Say it again, kid!” for L2 English for Finnish-speaking 6-9 year old children.