Research on Knowledge Graph Entity Recognition and Relation Extraction Algorithm Based on Deep Learning

This project intends to study deep learning methods for entity recognition and relationship extraction for knowledge graphs based on graph attention networks. First, this paper will establish a multi-layer graph model. The key nodes on the graph are dynamically adjusted using the attention mechanism between nodes. This enables it to better pay attention to the key nodes on the graph, thereby improving its extraction efficiency. Secondly, this project uses a self-supervised learning algorithm to optimize the model to further improve the generalization performance and accuracy of the model. This paper will use multiple public data sets to test the proposed algorithm and compare it with existing research results. Experiments have shown that the algorithm is better than the control model in terms of accuracy, recall rate and F1 value, especially for complex graph structure extraction. The method in this paper is about 12% higher than the traditional method on public data, and it has higher semantic understanding ability. The proposed method has strong adaptability and robustness for data of various complexities. This method can solve the knowledge graph problem well.