Abstract: 【Purpose/significance】In order to automatically recommend doctors to users according to their actual needs in the online medical community, this paper proposes a hybrid doctor recommendation algorithm based on similar users and similar doctors, accord⁃ ing to online consultation text information.【Method/process】Firstly, starting from the user counseling questions, find users which has similar counseling problems, and the doctors they selected are taken as the recommendation set based on similar users. Then from the doctor's answers, through LDA topic model training, dig out the implied disease theme from the doctor's answer text set, and find the doctor with similar disease diagnosis and treatment experience as a recommendation set according to the theme; Finally, the final rec⁃ ommendation list is obtained by combining the recommendation results of similar users and similar doctors through the calculation of mixed similarity.【Result/conclusion】The empirical results show that the method proposed in this paper can effectively reduce the da⁃ ta dimension, dig the potential semantic correlation between texts, effectively narrow the semantic gap, improve the recommendation quality, and have a good recommendation effect.【Innovation/limitation】This paper only selected a small sample data from the depart⁃ ments to conduct experiments, some parameters used empirical values. In the future, the application of this method to large-scale med⁃ ical data sets should be further explored.