Unsupervised entity alignment method and system between large-scale cross-language knowledge maps

The invention discloses an unsupervised entity alignment method and system between large-scale cross-language knowledge maps, and belongs to the field of entity alignment. In the feature embedding stage, based on multi-view information of a knowledge graph, embedding is carried out from two levels of semantics and characters. In the semantic level, a pre-trained large language model (LLM, Large Language Model) is used for embedding. And in a character level, an N-Gram model is used for embedding. In the alignment stage, firstly, multi-view information is fused, and similarity matrixes of semantic and character levels are generated respectively. Fusing the similarity matrix of the semantics and the characters in a weighting mode to generate an alignment matrix; and finally, completing an alignment task according to the alignment matrix. According to the method, any data annotation does not need to be carried out in advance, and unsupervised entity alignment is achieved; through feature embedding in two levels