基于孪生网络的中医医案主诉匹配方法

Matching Method for Main Complaint in Traditional Chinese Medicine Cases Based on Siamese Neural Network

针对中医医案匹配中文本语义关系交互性不足,不能充分利用文本间的语义信息来完成匹配任务和文本特征中噪音干扰难以去除使得计算精确度较低的问题,提出一种基于语义特征交互的中医医案主诉文本匹配模型,构建基于BERT模型的文本语义特征提取网络,提取句子中存在的语义特征信息,再利用向量交互的方式进行特征降噪和特征增强,经由分类器得到匹配结果。可以更好地契合中医医案实际任务场景,从而使匹配结果更准确有效。实验结果表明,所提出的方法对比其他对比模型具有更高的匹配准确率,是一种比较可行的解决方案。

Due to the lack of interactivity of Chinese text semantic relations in TCM medical case matching,the semantic information among texts cannot be fully used to complete the matching task,and the noise interference in text features is difficult to remove,which makes the calculation accuracy low,this paper proposes a TCM case main complaint text matching model based on semantic feature interaction.It constructs a text semantic feature extraction network based on the BERT model,extracts semantic feature information existing in sentences,and then uses vector interaction for feature denoising and enhancement.The matching results are obtained through a classifier.It can better fit the actual task scenarios of TCM medical cases,thereby making the matching results more accurate and effective.The experimental results show that the proposed method has higher matching accuracy compared to other comparative models,and is a relatively feasible solution.