文献类型：期刊文献

中文题名：基于深度强化学习的中医古籍图谱推理研究

英文题名：Study on the Inference of Knowledge Graph of TCM Ancient Books Based on Deep Reinforcement Learning

作者：刘悦悦[1];李燕[1];李春雨[1];刘雪丽[1]

第一作者：刘悦悦

机构：[1]甘肃中医药大学信息工程学院,甘肃兰州730000

第一机构：甘肃中医药大学信息工程学院（教育技术中心）

年份：2024

卷号：31

期号：6

起止页码：54

中文期刊名：中国中医药信息杂志

外文期刊名：Chinese Journal of Information on Traditional Chinese Medicine

收录：CSTPCD;;CSCD:【CSCD_E2023_2024】；

基金：甘肃省基层医疗卫生机构中医诊疗区(中医馆)健康信息平台(2305181101);中国高校产学研创新基金(2021LDA09002)。

语种：中文

中文关键词：知识图谱;知识推理;中医古籍;数据挖掘

外文关键词：knowledge graph;knowledge reasoning;TCM ancient books;data mining

摘要：目的针对中医古籍知识图谱实体关系信息缺失问题,提出基于深度强化学习的知识推理方法,提升知识图谱的完备性。方法以知识图谱领域基准数据集及中医古籍数据集为研究对象,采用基于深度强化学习推理方法,结合知识单步补全和多跳路径搜索,利用Python语言联合Neo4j图数据库实现路径间的推理。结果在知识推理任务数据集中,相较于基线模型,该推理方法在MAP指标上最高提升约53.79%,应用到中医古籍图谱时,其评价指标最高达0.776。结论基于深度强化学习的知识推理方法对于提升中医古籍知识图谱的完备性具有显著优势,该方法可有效地填补中医古籍图谱中的信息缺失。
Objective To address the issue of missing entity relationship information in the knowledge graph of TCM ancient books;To propose a knowledge reasoning method based on deep reinforcement learning;To improve the completeness of the knowledge graph.Methods Based on the benchmark dataset in the field of knowledge graph and the dataset of TCM ancient books,a deep reinforcement learning inference method was adopted.Combined with knowledge one-step completion and multi hop path search,Python language was used in conjunction with the Neo4j graph database to achieve path to path inference.Results In the knowledge reasoning task dataset,compared with the baseline model,this reasoning method improved the MAP index by about 53.79%.When applied to the knowledge graph of TCM ancient books,its evaluation index reached the highest of 0.776.Conclusion The knowledge reasoning method based on deep reinforcement learning has significant advantages in improving the completeness of the knowledge graph of TCM ancient books,and this method can effectively fill the information gap in the graph of TCM ancient books.