Large model assisted self-lifting multi-modal industrial equipment knowledge graph construction method

Abstract

The application discloses a large model assisted self-promotion multi-modal industrial equipment knowledge graph construction method, and particularly relates to the field of industrial equipment knowledge graph construction; through obtaining a multi-modal event fragment set, combining equipment topology snapshots and position number evolution information to generate the same object candidate set and the prohibited merging constraint set, using a large model to extract an entity relationship candidate set with modal evidence anchors, timestamps and version identifiers; further constructing an initial graph and generating a local counterfactual fragment set, obtaining an anti-confusion score by calculating the relationship establishment difference value and the constraint violation value, and screening to form a trusted subgraph; finally, using the trusted subgraph as a positive sample and the counterfactual negative relationship as a difficult negative sample to update the large model, and re-extracting and completing the graph for unconfirmed fragments; the method can effectively reduce the false association caused by position number replacement, topology proximity and time co-occurrence, and improve the accuracy and traceability of the knowledge graph.

Images