A Biomedical Relation Extraction Method Based on Graph Convolutional Network with Dependency Information Fusion

Biomedical texts are relatively obscure in describing relations between specialized entities, and the automatic extraction of drug–drug or drug–disease from massive biomedical presents a challenge faced by many researchers. To this end, paper designs relation method based on dependency information fusion to improve predictive power model for given entities. Firstly, we propose local–global pruning strategy syntax tree. Next, construction type matrix pruned tree incorporate sentence into feature extraction. We then attention mechanism graph convolutional calculating weights word–word dependencies, thus improving traditional network. The distinguishes importance different weights, weakening influence interfering such as word-to-word dependencies that unrelated entities long sentences. In paper, our proposed Dependency Information Fusion Attention Graph Convolutional Network (DIF-A-GCN) is evaluated two datasets, DDI CIVIC. experimental results show outperforms current state-of-the-art models.