Yinshang oracle bone inscriptions carry the genes of Chinese culture and contain rich musicological information. A classification model based on semantic relationship analysis is proposed for the related document data of Yinshang oracle bone music texts. The model selects document frequency and CHI statistics as the feature selection method, adopts vector space model for text representation, and applies KNN algorithm as the classification algorithm. Through experimental analysis on the dataset, the superior performance of the classification model in this paper is verified, and its average classification accuracy and recall rate reach 89.29% and 8.61%, respectively, which are higher than those of other classification methods with the combination of feature selection and classification algorithms. The classification system of oracle bone music can be categorized into four main types: ritual music, rain prayers, court feasts and court celebrations. In this paper, we refine the classification of oracle bone music rhetoric, which can help to provide practical references of rhetoric categories for oracle bone research, and provide retrieval convenience for related disciplines to utilize the oracle bone materials.
