Application of neural networks to braille transcription

We applied neural networks to wakachikaki, which is to transcribe Japanese text into Braille. The neural network with 3 layered structure was used and were trained by the method of back propagation. The highest average correct ratio of 94% was obtained by optimizing parameters in the neural network. This value seems to be high in the neural network without a large scale dictionary. Therefore, the application of the neural network into the wakachikaki is thought to be useful for obtaining high ratio of correct answer.