An on-line character recognition system for effective Japanese input

An online character recognition system has been developed in order to allow effective Japanese entry via a pen interface. The system consists of handwriting recognition, contextual postprocessing, and enhanced Kana-to-Kanji conversion subsystems. The recognition subsystem uses a hybrid algorithm consisting of a pattern matching discriminator and a neural network verifier. Special care was taken to improve recognition of non-Kanji and simple Kanji characters frequently used in fast data entry. The postprocessor predicts consecutive characters on the basis of biagrams modified by the addition of parts of speech and substitution of macro characters for Kanji characters. A Kana-to-Kanji conversion method designed for ease of use with a pen interface is also integrated into the system. In an experiment in which 2499 samples of Kanji and non-Kanji characters were taken from 20 subjects, it was observed that the original recognition accuracy of 85% (the result obtained by using the pattern matching discriminator) was improved to 91.5% by adding the hybrid algorithm, and that it was further improved to 94.5% by adding the postprocessor. Improved recognition accuracy.<<ETX>>