Intelligent dialogue question and answer method and system based on comparative learning

The invention provides an intelligent dialogue questioning and answering method and system based on comparative learning, and the method comprises the steps: inputting target question data into a semantic expression intelligent recognition model to output a semantic expression vector, performing fine tuning training on the pre-training model by adopting a sample space; the sample space comprises similar positive and negative sample pairs, dissimilar positive and negative sample pairs and negative samples generated based on a text generation type dialogue big language model; searching similar question vectors from the open source vector database, judging whether the similar question vectors contain target similar question vectors corresponding to the target question data or not on the basis of a threshold value, and if yes, outputting dialogue question and answer data. On the basis of reducing the calculation complexity, the accuracy and the real-time performance of question recognition and answer retrieval ca