Enhanced self-training-based medical scene dialogue generation method and system

The invention provides a medical scene dialogue generation method and system based on enhanced self-training, and relates to the technical field of large language models, and the specific scheme comprises the steps: obtaining a question text of a user in a medical scene; inputting a question text into the large language model after the self-training is enhanced, and generating an answer text through reasoning; wherein the enhanced self-training comprises the steps of forming an answer comparison group by using an answer text obtained by reasoning of a large language model and a text in which a standard answer text given by a medical expert is inconsistent, constructing a preference data set, and training the large language model by adopting a direct preference optimization algorithm; according to the method, the SFT and the enhanced self-training are combined, and the direct preference optimization algorithm is applied to the training process of the large language model, so that the reasoning accuracy of the