MGCN: Medical Relation Extraction Based on GCN

With the progress of society and improvement living standards, people pay more attention to personal health, WITMED (Wise Information Technology med) has occupied an important position. The relationship prediction work in medical field high requirements on interpretability method, but between entities is complex, existing methods are difficult meet requirements. This paper proposes a novel information relation extraction method MGCN, which combines contextual provide global for entities. uses Co-occurrence Graph Convolutional Network build up network relations entities, Open-world Assumption construct potential associated goes through Knowledge-aware Attention mechanism give entity pair interest. Experiments were conducted public dataset CTF, MGCN achieved score 0.831, demonstrating its effectiveness extraction.