Individual long-term memory hierarchical evolution and cognitive empowerment method and system

Abstract

The application discloses a kind of individual long-term memory layered evolution and cognitive empowerment method and system, belong to big language model memory management field.The method establishes behavior constraint layer, cognitive framework layer and key anchor layer, and sets stable cognitive track, exploration cognitive track and situational condition branch in cognitive framework layer;From the dialogue flow of user and big language model, extract candidate memory entry, determine parent stable cognitive framework item and current application situation, obtain semantic deviation degree, situation matching degree and stable verification information, execute first shunt according to preset order;State check is carried out to exploration cognitive framework item, and unified stable positioning processing is carried out to the item meeting stable verification condition;Memory entry is filtered according to dialogue type and evolution state and big language model input text is assembled.The application makes repetitive support, situation differentiation, cognitive update and to be verified content correspond to different data structure, evolution state and call permission.

Images