Mode-oriented non-functional demand knowledge refinement method

The invention discloses a mode-oriented non-functional demand knowledge refinement method, and relates to the field of demand engineering and knowledge graph refinement. The method comprises the following steps: expanding an established non-functional demand knowledge graph to obtain a quadruple set with a mode set; reading the quadruple set for preprocessing and initializing representation vectors of an entity, a relationship and a mode; projecting the vectors of the entities and the relationships to a mode space by using the projection matrix to obtain a scoring function; and detecting a hierarchical structure in the non-functional demand knowledge graph, calculating and obtaining a hierarchical constraint interval, obtaining a loss function in combination with a scoring function, carrying out optimization training on a knowledge graph embedding vector by adopting a gradient descent method to obtain a representation vector, and carrying out knowledge completion and knowledge debugging according to the repre