Research on Chinese medical entity relationship extraction based on multineural network and syntactic information

With the advancement of digitalization, the healthcare sector has accumulated vast textual data. This paper proposes a Chinese medical entity relationship extraction model that combines multiple neural networks and dependency information. By extending BERT's positional encoding for longer texts and employing a pruning strategy with local and global connections to generate adjacency matrices, the model uses graph convolutional networks to capture complex entity dependencies. Experimental results show that the proposed approach outperforms current baseline models on the TCM dataset, significantly improving relationship extraction performance.