Computational Analysis of Acquired Dyslexia of Kanji Characters Based on Conventional and Rough Neural Networks

Summary

Acquired dyslexia of Kanji characters is one of the most interesting research areas in neuropsychology. Although dyslexia of alphabets is known as the malfunction of the gyrus angularis (classical hypothesis), reading and writing of Chinese characters (Kanji characters) are intact in Japanese patients who suffered from cerebral infarction of the gyrus angularis. Thus, it has been pointed out that another group of neurons that integrates shape information should play an important role in recognition of Kanji characters (Iwata’s model). In this chapter, two neural network models, based on the classical hypothesis and Iwata’s model, are introduced to examine the validity of these two hypotheses. The computational experiments give the following three results: (i) Iwata’s model learns Kanji characters much faster than the classical model; (ii) Iwata’s model is more robust with respect to the malfunction of neurons; (iii) Iwata’s model simulates the characteristics of the two types of Japanese dyslexia.