Knowledge graph fusion method based on entity merging

• Main Authors: ZENG QINGXING, HAO ZHIFENG, HUANG HAN, ZHU HAOFENG
• Format: Patent
• Language: Chinese; English
• Published: 06.12.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a knowledge graph fusion method based on entity merging, which comprises the following steps of: firstly, acquiring data of two or more knowledge graphs to be fused, then extracting all attributes of entities, calculating word embedding vectors of the attributes, inputting the word embedding vectors into a recurrent neural network, acquiring final hidden layer output, and outputting the final hidden layer output. And finally, calculating the similarity of the entity attribute embedding vectors in the two maps by using cosine similarity, obtaining entities of two different maps which exceed a set threshold and have the highest similarity, and fusing the entities to obtain a fused knowledge map. According to the method, based on the recurrent neural network, the attributes of the entities are regarded as contexts, the entity embedded vectors in the two maps are mapped to the same feature space, the embedded vectors can contain all attribute information and the same dimension of the entit