CapsNet-based Chinese character representation model
Article from
Article citation format
Xie Haiwen, Ye Dongyi, Chen Zhaojiong.CapsNet-based Chinese character representation model [J] .Pattern Recognition and Artificial Intelligence, 2019,32 (02): 169-176.
The original download address: https://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&dbname=CJFDLAST2019&filename=MSSB201902009&v=MjkyNjJDVVI3cWZaZVpvRnk3blY3M0JLRDdZYkxHNEg5ak1YWTlGlVYG1V
1. Summary
A Chinese character glyph representation model based on CapsNet is proposed, and the Chinese character glyph representation is realized by characterizing the components in the Chinese character glyph.
(1) Generate characterization vectors of all component categories for any Chinese font ;
(2) According to the probability of component existence, using outlier detection based on Euclidean distance , select the corresponding component characterization vector;
(3) The glyph of the Chinese character is composed of the selected component representation vectors