Query extension optimization method and query method in open domain question and answer task

The invention discloses a query expansion optimization method and a query method in an open domain question and answer task. According to the query expansion optimization method, a three-step progressive prompt strategy of analysis, generation and refining is adopted, and query expansion generation of an open domain is promoted by utilizing the multi-task capability of LLMs; the method comprises the following steps: 1) in an analysis stage, generating analysis by utilizing semantic analysis capability of LLMs; 2) in a generation stage, generating various answer-oriented query extensions as candidates by utilizing knowledge retrieval and integration capability of LLMs, and retrieving a reference text to screen out wrong and irrelevant generation information and generate new candidates closer to potential answers; and 3) in order to perform quality control on the query expansion, performing self-review on all candidates by utilizing context understanding and reasoning capability of LLMs in a refining stage so a