ASR4PESP: Automatic Speech Recognition for Primary Education Speaking Proficiency

Duration November 2020 - November 2021  
PI dr. Catia Cucchiarini, Radboud University Nijmegen
Funders CITO, Clariah

Evaluating speaking proficiency in primary school is challenging as teachers have to judge many aspects (content and grammar) in one go. As a possible solution, CitoLab developed Beeldverhaal, an app that asks pupils to tell a story based on four pictures.

Through Automatic Speech Recognition the story is then converted into text, so that teachers can analyze the written text multiple times on multiple levels to judge the pupil’s level of speaking proficiency.
Since ASR of child speech is still problematic, the present project will try to improve its quality to make it suitable for educational applications like Beeldverhaal.