Abstract: Localization of zang-fu organs is a method to determine the location of lesions in zang-fu organs. It is an important stage of differentiation of zang-fu organs in traditional Chinese medicine(TCM). The purpose of this paper is to establish the localization model of TCM zang-fu organs through the neural network model. Through the input of symptom text information, the corresponding lesion zang-fu label can be output to provide support for the realization of zang-fu syndrome differentiation in TCM assisted diagnosis and treatment. In this paper, the localization of zang-fu organs is abstracted as multi-label text classification in natural language processing. Based on the medical record data of traditional Chinese medicine, a zang-fu localization model based on pre-training model ALBERT(A Lite BERT) and Bi-GRU (bidirectional gated circulation unit) was proposed. The F1-value of this model reached 0.8013 on the test set. The comparison and ablation experiments finally show that the proposed method is more accurate than multilayer perceptron and decision tree. And using ALBERT pre-training model for text representation effectively improves the accuracy of the model.