基于双向GRU和CNN的药物相互作用关系抽取

Extraction of drug-drug interaction based on bidirectional GRU and CNN

不同药物由于药效动力学和药代动力学的差异可能会产生不可预知的副作用,甚至威胁患者的生命安全。在信息技术飞速发展及指数级生物医学文献增加的背景下,从文本中提取药物相互作用成为可能,为此本文提出一种基于双向门控循环单元(GRU)和卷积神经网络(CNN)相融合的双层药物关系抽取模型,使用DDIExtraction2013作为数据集进行多组实验评估,实验结果获得最高75%的综合测评率;与其他方法相比较,基于双向GRU和CNN的双层模型可以有效地抽取文本中的药物相互作用关系。

Drug-drug interaction(DDI)is the difference between pharmacodynamics and pharmacokinetics of different drugs,which may produce unpredictable side effects or even threaten the life safety of patients.With the rapid development of information technology and the increase of exponential biomedical literature,it is possible to extract drug interactions from texts.A two-layer drug relationship extraction model based on the fusion of bidirectional GRU(bi-gated recurrent unit,BiGRU)and convolutional neural network(CNN)is proposed.DDIExtraction 2013 is used as data set to evaluate multiple groups of experiments,and obtain the highest comprehensive evaluation rate of 75%is obtained.Compared with other′s works,the two-layer model based on bidirectional GRU and CNN can effectively extract the drug interaction relationship in the text.