Prompt framework based on large language model and system and application thereof

The invention relates to the technical field of prompt engineering, in particular to a prompt framework based on a large language model and a system and application of the prompt framework, and the proposed prompt framework dynamically loads dominant prompt words and soft prompt parameters according to downstream task requirements so as to migrate the general capability of the large model to a task in a specific field. The multi-round dialogue implementation system adopts a pipelined architecture and comprises a natural language understanding module, a dialogue management module and a natural language generation module. In the implementation of multiple rounds of dialogues, the system executes tasks such as natural language understanding, dialogue management and natural language generation by using the architecture. The modules cooperate with each other to realize an efficient and coherent dialogue process. Experimental results show that through reasonable construction of cue words and dominant prompt, the mi