ReaDirect: An AI-Based Oral Reading and Comprehension Learning-Support Intervention Using Automatic Speech Recognition

This study developed and evaluated ReaDirect, an automatic speech recognition (ASR)-assisted oral reading and comprehension intervention aligned with the Philippine Department of Education ARAL Program. Thirty High Emerging Readers participated at one public elementary school: 10 Grade 1 learners in a feasibility component and 20 Grade 3 learners randomly assigned to experimental and control groups. The intervention combined a diagnostic assessment, six structured lessons, a final assessment, bounded feedback, and teacher monitoring. All Grade 1 learners completed the sequence, although eight required additional navigational assistance. In Grade 3, baseline-adjusted official English CRLA posttest results favored the experimental group for correct words, reading rate, and comprehension after Holm adjustment. Technical records showed reduced speech-recognition error rates and 88 passing targeted software tests. Findings support supervised supplementary use, but the small single-school sample and remaining technical validation gaps limit generalization. This preprint is a condensed report of the authors’ 2026 undergraduate thesis at Laguna State Polytechnic University. Each intervention learner completed one supervised ReaDirect session.