DART — Radboud University's Speech-Recognition Reading Tutor Boosts First-Graders' Decoding Skills in the Netherlands

DART (Dutch ASR-based Reading Tutor) is a reading-practice system built by the Centre for Language and Speech Technology (CLST) at Radboud University in Nijmegen, funded by the Netherlands Initiative for Education Research (NRO) since 2020. It uses automatic speech recognition to listen to young children read words and sentences aloud and gives them immediate, automated feedback, addressing the shortage of one-on-one reading practice time in regular classrooms.
A peer-reviewed study of 525 first-graders across 44 Dutch primary schools compared explicit-immediate versus implicit-delayed feedback over a six-week intervention. Explicit feedback produced significantly larger reading-accuracy gains on sentences (e.g. beta=-7.38, p<.001), and feedback that supplied the correct sound before children had to produce it themselves helped more than passive exposure. The researchers openly note a measurement limitation: the system's agreement with human teachers on reading errors was weaker for isolated words than for full sentences.
Since March 2023, DART's technology has been integrated into "Veilig leren lezen," a widely used Dutch reading-instruction method published by Zwijsen, with 1,200-1,400 pupils practising with the tool daily as of 2023 -- a rare case of a peer-reviewed speech-recognition reading tutor moving from a research RCT into mainstream commercial classroom use.

Read the full analysis: https://www.ru.nl/en/research/research-projects/dart-dutch-automatic-reading-tutor