Predicting oral reading fluency in Korean EFL learners using speech and reading comprehension measures*

Oral reading fluency (ORF) refers to reading a text quickly, accurately, and with appropriate phrasing and expression while comprehending its meaning. However, in schools, ORF assessment often focuses on reading speed and accuracy, despite the importance of prosody and reading comprehension. In addition, ORF is typically evaluated through manual teacher scoring, placing a substantial burden on teachers in large-scale assessments. As a foundational study toward automated ORF assessment, this study proposes and validates predictive models of teacher-rated segmental accuracy and prosody using manually derived speech and reading comprehension measures from Korean middle school English as a Foreign Language (EFL) learners. Students completed a 1-minute oral reading task, 4 passage comprehension items, and 14 silent reading comprehension items. Four English teachers rated the segmental accuracy and prosody on a five-point scale. Inter-rater reliability was examined using Krippendorff’s alpha, and Many-facet Rasch Measurement was used to yield the adjusted ORF scores. The analyzed speech measures included words per minute, words correct per minute, syllables correct per minute, syllables correct per second, accuracy, and F0 range. Most speech measures differed significantly across levels, except for the F0 range. Predictive models combining speech and silent reading comprehension measures outperformed other models, supporting their use in future automated ORF assessments.