基于神经网络的汉字多音字标音处理方法

A New Approach to the Grapheme-to-Phoneme Conversion

提出基于神经网络的多音字标音处理方法。该方法具有结构简单、易实现和对汉语知识依赖性不强的特点,而且通过开集测试和闭集测试,证明神经网络解决多音字标音处理具有较高的正确率。

A new approach to the Grapheme-to-phoneme conversion based on the artificial neural network is proposed in this paper. This approach is of simple structure, easy to be realized and does not require profound Chinese knowledge. Moreover, this new approach is of high accuracy, which has been proved by open and close tests.