Yu Bai(Radboud University), Ferdy Hubers(Radboud University), Catia Cucchiarini(Radboud University Nijmegen) and Helmer Strik(Centre for Language and Speech Technology (CLST), Centre for Language Studies (CLS), Radboud University Nijmegen)
Learning to read is a prerequisite to participate in our knowledge society. Developing reading skills requires intensive practice with individual evaluation and guidance by teachers, which is not always feasible in traditional classroom instruction. Automatic Speech Recognition (ASR) technology could offer a solution, but so far it has been mostly used to follow children while reading and to provide correct word forms through text-to-speech technology. However, ASR could possibly be employed at earlier stages of learning to read when children are still in the process of developing decoding skills. Early evaluation through ASR and individualized feedback could help achieve more personalized and possibly more effective guidance, thus preventing reading problems and improving the process of reading development.
In this paper we report on an explorative study in which an ASR-based system equipped with logging capabilities was developed and employed to evaluate decoding skills in Dutch first graders reading aloud, and to provide them with detailed, individualized feedback. The results indicate that ASR-based feedback leads to improved reading accuracy and speed and that the log-files provide useful information to enhance practice and feedback, thus paving the way for more personalized, technology-enriched approaches to reading instruction.