论文信息 - Training LSTM-RNN with Imperfect Transcription: Limitations and Outcomes

Training LSTM-RNN with Imperfect Transcription: Limitations and Outcomes

Bidirectional LSTM-RNN have become one of the standard methods for sequence learning, especially in the context of OCR due to its ability to process unsegmented data and its inherent statistical language modeling [5]. It has recently been shown that training LSTM-RNNs even with imperfect transcriptions can lead to improved transcription results [7, 14]. The statistical nature of the LSTM's inherent language modeling can compensate for some of the errors in the ground truth and learn the correct temporal relations. In this paper we systematically explore the limits of the LSTM's language modeling ability by comparing the impact of imperfect transcriptions with various hand crafted error types and real erroneous data created through segmentation and clustering. We show that training LSTM-RNN with imperfect transcriptions can produce useful OCR models even if the ground truth error is up to 20%. Further we show that it can compensate for some handcrafted error types with error rates of up to 40% almost perfectly.

Andreas Dengel | Syed Saqib Bukhari | Martin Jenckel

[1] Rob Fergus,et al. Learning from Noisy Labels with Deep Neural Networks , 2014, ICLR.

[2] Jürgen Schmidhuber,et al. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks , 2006, ICML.

[3] Nir Shavit,et al. Deep Learning is Robust to Massive Label Noise , 2017, ArXiv.

[4] Andreas Dengel,et al. OCRoRACT: A Sequence Learning OCR System Trained on Isolated Characters , 2016, 2016 12th IAPR Workshop on Document Analysis Systems (DAS).

[5] Jürgen Schmidhuber,et al. Long Short-Term Memory , 1997, Neural Computation.

[6] Klaus U. Schulz,et al. Automatic quality evaluation and (semi-) automatic improvement of mixed models for OCR on historical documents , 2016, ArXiv.

[7] Jürgen Schmidhuber,et al. Learning to Forget: Continual Prediction with LSTM , 2000, Neural Computation.

[8] Didier Stricker,et al. A comparison of 1D and 2D LSTM architectures for the recognition of handwritten Arabic , 2015, Electronic Imaging.

[9] Andreas Dengel,et al. anyOCR: A sequence learning based OCR system for unlabeled historical documents , 2016, 2016 23rd International Conference on Pattern Recognition (ICPR).

[10] Jürgen Schmidhuber,et al. Biologically Plausible Speech Recognition with LSTM Neural Nets , 2004, BioADIT.

[11] Thomas M. Breuel,et al. High-Performance OCR for Printed English and Fraktur Using LSTM Networks , 2013, 2013 12th International Conference on Document Analysis and Recognition.