Paper Info
Title
Sequence Discriminative Training for Offline Handwriting Recognition by an Interpolated CTC and Lattice-Free MMI Objective Function
Abstract
We study two sequence discriminative training criteria, i.e., Lattice-Free Maximum Mutual Information (LFMMI) and Connectionist Temporal Classification (CTC), for end-to-end training of Deep Bidirectional Long Short-Term Memory (DBLSTM) based character models of two offline English handwriting recognition systems with an input feature vector sequence extracted by Principal Component Analysis (PCA) and Convolutional Neural Network (CNN), respectively. We observe that refining CTC-trained PCA-DBLSTM model with an interpolated CTC and LFMMI objective function ("CTC+LFMMI") for several additional iterations achieves a relative Word Error Rate (WER) reduction of 24.6% and 13.9% on the public IAM test set and an in-house E2E test set, respectively. For a much better CTC-trained CNN-DBLSTM system, the proposed "CTC+LFMMI" method achieves a relative WER reduction of 19.6% and 8.3% on the above two test sets, respectively.
Year
DOI
Venue
2017
10.1109/ICDAR.2017.19
2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)
Keywords
Field
DocType
Lattice-Free MMI,CTC,Sequence Discriminative Training,Offline Handwriting Recognition
Feature vector,Mathematical optimization,Pattern recognition,Convolutional neural network,Computer science,Word error rate,Handwriting recognition,Feature extraction,Artificial intelligence,Mutual information,Discriminative model,Test set
Conference
Volume
ISSN
ISBN
01
1520-5363
978-1-5386-3587-2
Citations 
PageRank 
References 
1
0.37
14
Authors
8
Name
Order
Citations
PageRank
Wenping Hu1826.77
Meng Cai2688.24
Kai Chen3715.38
Haisong Ding492.57
Lei Sun5183.40
Sen Liang681.21
Xiongjian Mo730.77
Qiang Huo8109899.69