Word-class bigram statistics language model for a hand-written chinese character recongizer

摘要

A word-class bigram statistics language model supported by a huge Chinese lexicon of 87, 326 word entries has been investigated on its effectiveness in upgrading the accuracy of a hand-written Chinese character recongnizer. The concept of the lhogeneity of a word-class is introduced in classifying words in to word-classes. On the average, the bigram statistics language model upgrades the recongition rate by 12.4