Purpose/Significance Mining the medical characteristics of online hospitals plays a very important role in online medical accurate recommendation. At present, although some online hospitals have the feature label function, they can only realize the internal feature prompt, and can not measure the feature differences between different hospitals from a global perspective. Method/Process The paper proposes a hospital special medical based LDA (HSM_LDA) model based on online hospital inquiry text. The method takes hospital ID as the modeling entry, converts the “text-vocabulary” matrix in the inquiry corpus into the “hospital-vocabulary” matrix, jointly models the three variables of hospital, topic and vocabulary, and generates two distributions, “hospital-topic”(E)and “topic-vocabulary”(F). Finally, the medical characteristics of each hospital are identified by combining E and F distribution. Result/Conclusion The hospital consultation text provided by the “haodf.com” platform is used as the experimental data set, and the HSM_LDA model is used for mining and analysis. The experimental results show that the hospital characteristic recognition accuracy of the proposed method is 87%, which has achieved a good effect.