Semantic analysis of the Oracle Radicals using Similarity Strategy based on the Yolov2 network

Abstract

Oracle bone inscriptions are the earliest character system in China, and its pattern is similar to the hieroglyphic. This paper selects 14 classes, including six different font evolution forms: oracle bone inscriptions, bronze inscriptions, states inscriptions, seal inscriptions, official scripts and regular scripts. Through the analysis of the combination of oracle radicals and the heads of oracle bones, and use convolutional neural networks to analyze the evolution of the oracle bones structure of the evolution process. In this experiment, this paper uses Yolov2 model to achieve oracle radical detection, and the results show that the oracle radical detection accuracy can reach 95.5%, meanwhile, we prove that some oracle radicals have similar structure relationships between the evolution of characters.