Singular Value Decomposition is a Valid Predictor of Stroke Importance in Reading Chinese

Hsueh-Cheng (Nick) Wang
Department of Computer Science
University of Massachusetts at Boston

Chinese characters are written in a specific stroke order which 
may reflect a stroke’s importance. Previous research showed 
removing the initial strokes from characters made them harder 
to read than removing the final strokes or the shortest strokes. 
Singular value decomposition (SVD) may be a better way to 
estimate which elements of a character are most important for 
identification. This study decomposed characters into segments 
(i.e., vertical, horizontal, and diagonal lines), and the 
importance of each segment was determined by SVD. Subjects read 
50 sentences either with all or 70% of the segments retained. 
The retained segments were the most important, the least 
important, or randomly selected. When the most important 
segments were retained, subjects read as fast as when all were 
retained. When the least important segments were retained, 
reading fluency was reduced. The results suggest that SVD is a 
psychologically valid way to capture Chinese character 
configuration.