Research Aim
Early detection of reading difficulties is important for educational and developmental support. However, conventional oral reading tests and reading comprehension tasks do not always reveal where a child struggled or what kind of reading behavior occurred.
In this research, we record gaze data during oral reading and extract features such as reading speed, regressions, skipped lines, and fixation time. Machine learning is then used to help clinicians and supporters understand reading behaviors objectively.
Eye position and pupil information are measured during oral reading to visualize the reading process over time.
Reading speed, regressions, skipped lines, fixation bias, and other gaze features are extracted from time-series data.
Machine learning and explainable AI provide information that can support screening and intervention planning.
Main Topics
- Assessment of oral reading and reading comprehension using gaze data
- Extraction of gaze features such as regressions, skipped lines, and fixation time
- Machine-learning-based screening support for reading difficulties
- Visualization of decision factors using explainable AI methods such as SHAP
- Design of assessment reports for clinicians and supporters