Open knowledge graph construction method for cross-text potential fact discovery and anti-fact correction

Abstract

An open knowledge graph construction method for cross-text potential fact discovery and anti-fact correction belongs to the technical field of natural language processing and knowledge engineering, and comprises the following steps: firstly, extracting a triple from a text by using a large language model and defining a mode; secondly, aligning and normalizing the modes through semantic similarity calculation and threshold judgment, and constructing a unified semantic base; thirdly, aggregating cross-text evidences based on a clue retrieval strategy, and constraining a large language model to infer potential facts so as to enhance graph connectivity; finally, an anti-fact comparison mechanism is introduced, candidate facts are automatically corrected through confidence coefficient comparison, and the knowledge purity is improved. According to the method, knowledge fragmentation can be effectively relieved without retraining a basic model, fact errors are reduced, and a more complete, consistent and reliable open knowledge graph is constructed.

Images