国家重点研发计划课题(2019YFC1712301); 江西省教育厅科技技术研究重点项目(GJJ201204);江西省教育厅科学技术研究项目(GJJ170727);江西中医药大学博士启动基金(2018WBZR021); 江西省一流学科建设科研启动基金专项项目(JXSYLXK-ZHYI059)
Jiangxi University of Chinese Medicine
National Key Research and Development Project (2019YFC1712301); Jiangxi Provincial Department of Education Science and Technology Research Key Project (GJJ201204); Jiangxi Provincial Department of Education Science and Technology Research Project (GJJ170727); Jiangxi University of Traditional Chinese Medicine Doctoral Funding (2018WBZR021); Jiangxi Province's first-class discipline construction Special Project of Research Startup Fund (JXSYLXK-ZHYI059)
Chinese Medicine syndrome differentiation is the basis and premise of Chinese Medicine clinical legislation, prescription and medication. Due to the lack of high-quality corpus of electronic medical records of Chinese Medicine, the model training is prone to underfitting, and the symptom expression forms of the four diagnostic information are quite different, which limits the network model's ability to recognize complex symptoms. Aiming at the above problems, a multi-channel Chinese Medicine syndrome differentiation model integrating knowledge graphs is proposed by separately processing the multi-channels of the four diagnostic information and enhancing the knowledge of the model training by artificially constructed small-scale knowledge graphs. The experimental results show that, based on the Chinese Medicine electronic medical record data set, the proposed model is 3.51%, 3.38%, and 3.32% higher than the baseline model in terms of P@1 index, P@3 index, and P@5 index, respectively, compared with other network structures. The model has also been improved to varying degrees, which verifies that the proposed model has a significant effect on Chinese Medicine syndrome differentiation.
叶青,张素华,程春雷,等. 融合知识图谱的多通道中医辨证模型[J]. 科学技术与工程, , ():复制