Multi-Class Character Classification with Semi-Supervised Learning Based on Information Entropy

Abstract:

Character Classification technology is the key link in OCR system. Most classification methods require abundant marked samples training to get classifier. In the real OCR application, there are so many classes, to label these samples are often waste time and energy, especially for unacquainted language, such as Arabic and Uygur, many characters are difficult to differentiate, so it even needs the help of professional guidance. This paper proposed a novel character classification with semi-supervised learning based on information entropy, introduced discrete event probability estimation theory of information entropy, active to select the optimization character samples, got the new parameters to train the classifier again, choose the most conducive to the classifier performance samples, iteration until the unlabeled samples set is empty. The experiment results show that this method achieves high performance in specific condition.

• Periodical: Applied Mechanics and Materials (Volumes 347-350)
• Pages: 3167-3171
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.347-350.3167
• Online since: August 2013
• Authors: Xue Wei Wang, Yuan Yan Tang, Jia Duan, Liang Rui Peng, Zhen Chao Zhang
• Keywords: Arabic OCR, Character Classification, Information Entropy, Semi-Supervised

References

[1] Kamal Nigam, Andrew McCallum, Sebastian Thrun, and Tom Mitchell. Learning to classify text from labeled and unlabeled documents, In AAAI-98, (1998).

DOI: 10.21236/ada350490

[2] Xiaoqing Ding, and Youshou Wu, Information Entropy Theory in Pattern Recognition, ACTA ELECTRONICA SINICA, pages 21(8): 1-8, (1993).

[3] A.P. Engelbrecht, Incremental learning using sensitivity analysis, In Proceedings of International Joint Conference on Neural Networks, pages 1350-1355, July, (1999).

[4] CA Thompson, and RJ Mooney, Active Learning for Natural Language Parsing and Information Extraction, In Proceedings of sixteenth Internatlonal Machine Leaning Conference, Slovenia, (1999).

[5] DA Cohn, Z Ghahramani, and MI Jordan, Active learning with statistical models, Articial Intelligence Research, papers 4: 129-145, (1996).

DOI: 10.1613/jair.295

[6] R Liere, and P Tadepalli, Active learning with committees for text categorization, In Proceedings of the Fourteenth National Conference on Articial Intelligence, papers 591-596, (1996).

[7] A McCallum, and K Nigam, Employing EM in pool-based active learning for text classification, In Proceedings of the Fiteenth International Conference(ICML), papers 359-367, (1988).

[8] Chen Jinhe, Shenjie, Active L earning Based on Information Entropy for Semi-supervised Classification, Computer Technology and Development, 20(2), (2010).