caa a22 4500 44537957X CHVBK 20180317143013.0 cr unu---uuuuu 170323e20110901xx s 000 0 eng 10.1007/s10032-011-0160-x doi (NATIONALLICENCE)springer-10.1007/s10032-011-0160-x Confidence- and margin-based MMI/MPE discriminative training for off-line handwriting recognition [Elektronische Daten] [Philippe Dreuw, Georg Heigold, Hermann Ney] We present a novel confidence- and margin-based discriminative training approach for model adaptation of a hidden Markov model (HMM)-based handwriting recognition system to handle different handwriting styles and their variations. Most current approaches are maximum-likelihood (ML) trained HMM systems and try to adapt their models to different writing styles using writer adaptive training, unsupervised clustering, or additional writer-specific data. Here, discriminative training based on the maximum mutual information (MMI) and minimum phone error (MPE) criteria are used to train writer-independent handwriting models. For model adaptation during decoding, an unsupervised confidence-based discriminative training on a word and frame level within a two-pass decoding process is proposed. The proposed methods are evaluated for closed-vocabulary isolated handwritten word recognition on the IFN/ENIT Arabic handwriting database, where the word error rate is decreased by 33% relative compared to a ML trained baseline system. On the large-vocabulary line recognition task of the IAM English handwriting database, the word error rate is decreased by 25% relative. Springer-Verlag, 2011 Handwriting recognition nationallicence Arabic nationallicence Discriminative training nationallicence Maximum mutual information nationallicence Minimum phone error nationallicence Margin nationallicence Confidences nationallicence Hidden Markov models nationallicence Model adaptation nationallicence Dreuw Philippe RWTH Aachen University, Human Language Technology and Pattern Recognition, Ahornstr 55, 52056, Aachen, Germany aut Heigold Georg RWTH Aachen University, Human Language Technology and Pattern Recognition, Ahornstr 55, 52056, Aachen, Germany aut Ney Hermann RWTH Aachen University, Human Language Technology and Pattern Recognition, Ahornstr 55, 52056, Aachen, Germany aut International Journal on Document Analysis and Recognition (IJDAR) Springer-Verlag 14/3(2011-09-01), 273-288 1433-2833 14:3<273 2011 14 10032 https://doi.org/10.1007/s10032-011-0160-x text/html Onlinezugriff via DOI 1 research-article jats NATIONALLICENCE 856 40 https://doi.org/10.1007/s10032-011-0160-x text/html Onlinezugriff via DOI NATIONALLICENCE 700 1- Dreuw Philippe RWTH Aachen University, Human Language Technology and Pattern Recognition, Ahornstr 55, 52056, Aachen, Germany aut NATIONALLICENCE 700 1- Heigold Georg RWTH Aachen University, Human Language Technology and Pattern Recognition, Ahornstr 55, 52056, Aachen, Germany aut NATIONALLICENCE 700 1- Ney Hermann RWTH Aachen University, Human Language Technology and Pattern Recognition, Ahornstr 55, 52056, Aachen, Germany aut NATIONALLICENCE 773 0- International Journal on Document Analysis and Recognition (IJDAR) Springer-Verlag 14/3(2011-09-01), 273-288 1433-2833 14:3<273 2011 14 10032 Metadata rights reserved Springer special CC-BY-NC licence nationallicence BK010053 XK010053 XK010000 NATIONALLICENCE NATIONALLICENCE NL-springer