基于构词学变体的跨领域表型概念标准化方法  

Word-building variants-based normalization method for cross domain phenotype concepts

作　　者：傅筱 韩俊毅 曹阔 FU Xiao;HAN Jun-yi;CAO Kuo(Shanghai Lixin Accounting and Finance College Information Management School,Shanghai 201209,China;Chongqing Yingkezhushu Network Science and Technology Limited Company,Chongqing 401147,China)

机构地区：[1]上海立信会计金融学院信息管理学院,上海201209 [2]重庆英科铸数网络科技有限公司,重庆401147

出　　处：《中华医学图书情报杂志》2020年第6期42-48,共7页Chinese Journal of Medical Library and Information Science

摘　　要：标准化的医学术语对医生提供个性化治疗和患者掌握自身健康状况非常重要,一种适用于跨领域实体标准化的方法可以保证语料库管理的一致性,并减少歧义。分析了临床与生物医学领域的概念特征,并通过基于英语与希腊语构词学的语义转换提出了一种将不同领域的表型概念与术语库自动匹配的混合标准化方法。其匹配临床领域与生物医学领域表型概念的F 1值分别为0.7189与0.8366。与基于字典查询的检索方法相比,提升了4.95%与4.38%,可以有效完成跨领域的标准化概念识别任务,并具有良好的鲁棒性与扩展性,同时对计算机辅助的概念标注及连接中文概念与英文标准术语库的跨语言自动匹配具有一定的借鉴价值。Normalized medical terms are very important for physicians and surgeons to provide their personalized treatment of patients and for patients to know their health condition.A normalized method for cross domain entity can ensure the consistency of language corpus repository management and reduce its ambiguity.The characteristics of phenotype concepts in clinical and biomedical fields were analyzed,and a normalized method was proposed for automatic match of phenotype concepts and language corpus in different domains based on the semantic transformation of word-building in English and Greek,which showed that the F 1 value was 0.7189 and 0.8366 respectively for the matched phenotype concepts in clinical and biomedical fields,and was 4.95%and 4.38%higher than those of dictionary look up-based retrieval method,indicating that the normalized method for cross domain entity can effectively identify the phenotype concepts in cross domain field with a good robustness and a good expansibility,and is thus of a certain reference value for annotating the computer-assisted phenotype concepts and linking the cross language automatic match of Chinese concepts and English normal terms repository.

关 键 词：跨领域表型概念标准化 概念变体 构词学同义词 希腊语派生词 信息检索 

分 类 号：G254[文化科学—图书馆学] R-05[医药卫生]