Abstract: 【Purpose/significance】To reveal the potential semantic relationship between diseases in Chinese electronic medical re⁃ cords and solve the problem of semantic relationship ambiguity.【Method/process】In this paper, we construct an inter-disease rela⁃ tionship mining model based on entity recognition and co-occurrence analysis, and conduct an empirical study on an open electronic medical record dataset. For entity recognition, this paper mainly uses the BERT-BiLSTM-CRF deep model to extract diseases and re⁃ lated information from Chinese electronic medical records, uses the co-occurrence analysis method to quantify the semantic relation⁃ ship between diseases, and finally uses similarity calculation and hierarchical clustering to mine the semantic relationship between dis⁃ eases.【Result/conclusion】The deep learning model for named entity recognition has good performance, and the F1 score on the vali⁃ dation set reaches 0.95. The method of co-occurrence analysis can better mine the semantic relationship between diseases.【Innova⁃ tion/limitation】This paper fuses direct co-occurrence and indirect co-occurrence, and proposes a method based on comprehensive co-occurrence.