Neural network-based knowledge graph entity linking method and apparatus, and electronic device

The invention provides a knowledge graph entity linking method and device based on a neural network and electronic equipment, and aims to link an entity reference necklace in a document to a target entity of a given knowledge graph and can be used for construction and expansion of the knowledge graph. Entity reference items and entity contexts are coded by adopting a recurrent neural network by utilizing contexts and relationships of entities in a knowledge graph and graph structure relationships of neighbor entities and introducing relationship-entity pairs with importance marks of target entities; and establishing a correlation model between an entity reference item context and an entity context by using a common attention mechanism, modeling an entity by using a graph convolutional neural network, generating a comprehensive neural network model to process a text where the entity is located, and obtaining a knowledge graph entity link. In this way, the joint attention mechanism and the graph convolutional n