Minimum-risk training for semi-Markov conditional random fields with application to handwritten Chinese/Japanese text recognition
作者：
Highlights：
• We present a minimum-risk (MR) training method for semi-CRFs.
• MR aims at minimizing the character error rate rather than the string error rate.
• Three non-0/1 cost functions are compared with the conventional 0/1 cost.
• Lattice edge selection is investigated in MR to reduce the training complexity.
• Six learning criteria are evaluated on handwritten Chinese text recognition tasks.
摘要
Highlights•We present a minimum-risk (MR) training method for semi-CRFs.•MR aims at minimizing the character error rate rather than the string error rate.•Three non-0/1 cost functions are compared with the conventional 0/1 cost.•Lattice edge selection is investigated in MR to reduce the training complexity.•Six learning criteria are evaluated on handwritten Chinese text recognition tasks.
论文关键词：Semi-Markov conditional random fields,Minimum-risk training,Character string recognition
论文评审过程：Received 6 July 2013, Revised 24 October 2013, Accepted 2 December 2013, Available online 12 December 2013.
论文官网地址：https://doi.org/10.1016/j.patcog.2013.12.002
