采用混合方法抽取生物医学实体间语义关系  

Extract Semantic Relations Between Biomedical Entities Applied Hybrid Method

作　　者：王秀艳[1] 崔雷[2] 

机构地区：[1]南方医科大学珠江医院图书馆,广州510280 [2]中国医科大学信息管理与信息系统(医学)系,沈阳110001

出　　处：《现代图书情报技术》2013年第3期77-82,共6页New Technology of Library and Information Service

基　　金：国家自然科学基金项目"运用文本数据库中元数据关联规则进行知识发现的研究"(项目编号:70473101)的研究成果之一

摘　　要：以阿司匹林副作用为研究样本的主题,搜集PubMed数据库中与"药物副作用引起疾病"主题词关联规则相对应的文献记录,在文献摘要中提取出相应实体共现的句子,进而利用自然语言处理方法提取文献中表达生物医学实体间语义关系的动词,共提取出30个表示药物副作用与疾病关系的动词。研究结果表明:基于共现和自然语言处理提取语义关系动词的方法是可行的。This paper takes aspirin side effects as the research topic, and applies one established association rule （IF Drugs and Chemicals categories/adverse effects AND Diseases categories/chemically induced, THEN Drugs cause Disea- ses） to collect the bibliographic records in PubMed involved the MeSH association rules. Then, it extracts the co - occur- rence sentences that include the targeted entities and the semantic verbs between biomedical entities by the natural lan- guage processing method. Finally, 30 semantic verbs describing the relations between drug side effects and diseases are extracted. The research result shows that it is feasible to extract semantic relations based on the co - occurrence and natu- ral language processing methods

关 键 词：生物医学实体 语义关系抽取 自然语言处理 

分 类 号：G350[文化科学—情报学]