The introductory chapter describes the problem, gives an overview of the study of sound symbolism, describes the language corpus used, analyzes language variants in Japanese, and explains the terminology, transcription, and data sources used. We propose a new method of giving a certain meaning to each node of the hidden layer by introducing a dictionary of semantic vector words created manually with source symbols and using vector graphic models. The most reliable way to learn kanji for people living in Japan is to focus on learning kanji in order based on how often it is used in Japanese texts. It is calculated with the help of Japanese dictionaries for students. However, it is assumed that the same approach does not apply to people learning Japanese outside of Japan, due to the lack of a universal order in which kanji should be taught. The table shows the frequency of use of 2136 standardized Kanji yoosyoos. This study examines symbolic or mimetic sound words in Japanese, most of which have never been included in Japanese dictionaries and may be incomprehensible in all nuances to native speakers.