计算机科学与探索 ›› 2018, Vol. 12 ›› Issue (9): 1475-1486.DOI: 10.3778/j.issn.1673-9418.1806024

• 人工智能与模式识别 • 上一篇    下一篇

面向患者的智能医生框架研究

谢  刚1,2,吴高巍1,任俊宏1,张似衡1,牛景昊1,张文生1+   

  1. 1. 中国科学院 自动化研究所,北京 100080
    2. 贵州师范大学 大数据与计算机科学学院,贵阳 550001
  • 出版日期:2018-09-01 发布日期:2018-09-10

Research on Intelligent Doctor Framework for Patient

XIE Gang1,2, WU Gaowei1, REN Junhong1, ZHANG Siheng1, NIU Jinghao1, ZHANG Wensheng1+   

  1. 1. Institute of Automation, Chinese Academy of Sciences, Beijing 100080, China
    2. School of Big Data and Computer Science, Guizhou Normal University, Guiyang 550001, China
  • Online:2018-09-01 Published:2018-09-10

摘要: 目前,国内对中文智能医生的研究相对较少,针对患者需求的智能医生研究更少。鉴于此,提出了一种能及时准确地回答患者健康问题的“一问一答”智能医生框架。在该框架中,首先构建一个包含1 126 214个三元组的中文医学知识图谱和一个包含60万记录的问答库;其次提出一种基于依存关系的问题分析算法,以对用户的提问进行分析;再次提出一种将自然语言表述的临床表型数据转换成语义三元组的方法;然后提出一种问题综合评分算法对候选答案对应的问题进行评分;最后实现了一个妇产科智能医生原型系统,并利用真实的问答语料对其进行人工测试,实验结果表明使用该框架构建的智能医生框架可以有效地回答用户的问题。该项成果已成功应用于某公司的健康咨询APP中。

关键词: 智能医生, 智能问答, 知识图谱

Abstract: At present, there are few studies about intelligent doctor in China. Therefore, a Chinese intelligent doctor framework is proposed. It works by answering one question at a time and provides real-time and automatic question- answer service. Firstly, a Chinese medical knowledge graph which contains 1126214 statements about entities is constructed and a database about frequently asked questions which contains 600 thousand records is built. Secondly, a question analysis method based on dependency relation is proposed to analyze the question. Thirdly, a technique is developed to convert clinical phenotypic data stated in natural language into some semantic triples. Fourthly, an algorithm for scoring of the questions is proposed. Finally, an intelligent doctor of obstetrics and gynecology is constructed and the real data are used to test the doctor. The experiment results show that the framework is of effectiveness. The work has been embedded into an APP of a health consultant firm.

Key words: intelligent doctor, intelligent question-answer, knowledge graph