A Knowledge Graph Embedding Model Fusing With Entity Type Features And Its Application

With the explosion of Internet data,how efficiently obtain information and knowledge in data is a huge challenge.Now we tend to learn the knowledge in the data through the related knowledge acquisition models,and the accuracy of knowledge graph embeddings has an important fundamental impact on the performance of related knowledge acquisition models and is one of the research hotspots in the field of knowledge graphs.So this paper conducts theoretical and experimental research to improve the accuracy of knowledge graph embeddings.Firstly,since entity type information is often missed in the knowledge graph,this paper proposes an entity type feature extraction method based on the neighborhood relationship representation vector to achieve the goal of improving the model performance of knowledge acquisition tasks.The method makes full use of the internal structure information of the knowledge graph,which means that it fuses the entity type features into the embeddings of the knowledge graph entities and relations,and enhances the representation capability of the embeddings.The method firstly characterizes the entities based on neighborhood relations and then uses a combination of clustering and dimensionality reduction to extract the common features of each cluster entity and use them as entity-type features.Secondly,to make entity and relation embeddings contain richer semantic information and enhance the accuracy of embeddings,this paper proposes a knowledge graph embedding model(FTKGE)to enhance the accuracy of embedded representation.The model uses the method of entity type feature extraction based on a neighborhood relationship vector to extract entity type features.The entities and relations in the knowledge graph are represented and learned by the graph attention network,and the entity embeddings containing certain semantic information are obtained.The entity type features are fused with the entity embeddings to obtain more accurate entity embeddings.Finally,to verify the correctness of the proposed method,link prediction experiments are carried out on two data sets: WN18 RR and FB15k-237.And to reflect the progressiveness of the proposed method,this paper makes a comparative analysis with several advanced knowledge graph embedding models.The experimental results verify the effectiveness of the FTKGE model and entity-type feature extraction method.