【摘要】 甲状腺病史作为一类重要的非结构化文档,对医疗诊断至关重要。针对具体的甲状腺病史数据,提出一种基于深度学习的甲状腺病史结构化处理方法。首先,构建专业词库和病史本体,使用专业词库指导分词,基于本体结构完成结构化输出;其次,通过使用实体识别技术,完成对分词结果标签的预测;最后,使用标签抽取和词库匹配两种方法对病史数据进行信息抽取,并将结构化结果以RDF进行存储。实验结果表明该方法的准确率和泛化性较传统方法有明显提升。更多还原

【Abstract】 As an important class of unstructured documents,the medical record of thyroid disease is critical to disease diagnosis.According to the specific medical record of thyroid disease data,a method for structuring medical record of thyroid disease based on deep learning is proposed. Firstly,professional dictionary and medical record ontology are constructed. Thus,word segmentation is realized by using professional dictionary and structured output is completed based on ontology structure. secondly,entity recognition technology is employed to complete prediction of segmentation result label; finally,label extraction and dictionary matching are used to extract information from the medical record data,and the structured results are stored in RDF. The results of experiments showthat the accuracy and generalization of the method are significantly improved compared with the traditional methods.更多还原