基于BiLSTM-CRF的《神农本草经》命名实体识别研究OA
Research on Named Entity Recognition of Shen Nong's Materia Medica Based on BiLSTM-CRF
目的:基于BiLSTM-CRF的命名实体识别技术挖掘并展示《神农本草经》蕴含的药物理论.方法:构建自定义中医术语词库,由计算机自动化序列标注,根据不同主流命名实体识别方法以及中医古籍的文本特点,以字向量作为初始输入,构建BiLSTM-CRF模型对《神农本草经》进行命名实体识别.结果:测试结果表明,BiLSTM-CRF模型的精确率89.00％,召回率88.83％,F 1值为88.91％,相对于其他模型效果较优.结论:BiLSTM-CRF模型能够有效识别《神农本草经》的实体类型,适用于中医古籍的知识挖掘,有助于中医理论实践和发挥临床应用价值.
Objective:To mine and demonstrate the drug theories embedded in Shen Nong's Materia Medica based on BiLSTM-CRF named entity recognition technology.Methods:Build a custom vocabulary of traditional Chinese medicine terminology,annotated by computer automated sequences,according to different mainstream named entity recognition methods and the text characteristics of ancient Chinese medical texts,a BiLSTM-CRF model is constructed for named entity recognition of Shen Nong's Materia Medica with word vectors as the initial input.Results:The test results show that the precision of BiLSTM-CRF model is 89.00％,the recall rate is 88.83％,and the F1 value is 88.91％,which is better than other models.Conclusion:BiLSTM-CRF model can effectively identify the entity types of Shen Nong's Materia Medica,which is suitable for knowledge mining of ancient Chinese medical texts and helps to practice the theory of Chinese medicine and bring into play the value of clinical application.
周嘉玮;王坤;吴雨璐;李荣耀;刘秀峰
广州中医药大学医学信息工程学院,广东广州 510006||暨南大学信息科学技术学院,广东广州 511443广州中医药大学医学信息工程学院,广东广州 510006广州中医药大学医学信息工程学院,广东广州 510006||广州中医药大学智能研究院,广东广州 510006
药学
命名实体识别神农本草经中医古籍BiLSTM-CRF
Named entity recognitionShennong's materia medicaTraditional Chinese medicineBiLSTM-CRF
《成都中医药大学学报》 2024 (003)
54-59 / 6
国家级大学生创新创业训练计划项目(202310572031X);广东省大学生创新创业训练计划项目(202210572078);广东省普通高校特色创新类项目(2022WTSCX010)
评论