[2K4-GS-9] Human interface

[2K4-GS-9-04] Estimation of program comprehension using keystroke behavior of typing software

〇toyohisa nakada¹, motoki miura² (1. Niigata University of International and Information Studies, 2. Chiba Institute of Technology)

Keywords:key stroke, programming, typing, machine learning

The purpose of this research is to estimate from the behavior of keystrokes whether the user understands or does not understand the program when typing a predetermined program such as typing software. As a method, a one-dimensional convolutional neural network is used, and the input is normalized keystroke interval information, and the output is a binary value that indicates whether the program was understood. For comparison, we also created a model that linearly separates only by typing speed. As experimental data, we used the typing data and comprehension test results of 112 persons on university lectures. As a result, the comprehensibility of the program was estimated with an accuracy of about 70% for the simple linear separable model and about 80% for the proposed one-dimensional convolutional neural network.