Analyzing the impact of structure in handwriting recognition software

Abstract

Handwriting recognition for text and numbers has existed for Tablet PCs and specialty devices for almost a decade. The use of such software in a classroom can shorten the latency between answers to exercises and teacher feedback. While text and number inputs have already been well explored, graphical examples and math problems are a relatively new territory for recognition software. Under the guidance of the NSFfunded INK-12 Project, I explored the impact of structure on the ability of Ink-Analysis Handwriting Recognition to understand students' solutions math and science exercises, hoping to find the ideal balance between accurate grading and minimal human effort. I tested a prototype system aimed at supporinting "virtual-teacher-guided" learning in elementary school classrooms in the Greater Boston area.