A method for generating adaptive heuristic teaching dialogue based on MCP multidimensional context

Abstract

The application relates to the technical field of artificial intelligence education, and discloses a self-adaptive heuristic teaching dialogue generation method based on MCP multidimensional context. After receiving a student end teaching interaction request, the current state information of the student is pulled through an MCP protocol, including learning node coding, mastery probability score and mistake distribution; the learning node coding and the mistake distribution are taken as search keys to search for error-prone point analysis and guiding skill strategy in a teaching and research knowledge base; the state information, analysis data, skill strategy and concept hard boundary are spliced into system prompt words, a heuristic reply is generated in a concept hard boundary by inputting a teaching method fine-tuning large model; after verifying that the student passes the current guiding step, a mastery degree update signal is sent to a state server through the MCP protocol. The application fuses state perception and hard boundary constraint, and avoids directly giving answers and knowledge out of class.

Images