命名实体识别-北京邮电大学学报.PDF

文章编号：1007-5321(2006)05-0079-05 特定领域下关系模板的研究 张素香1,2， 李 蕾1， 谭咏梅1 (1．北京邮电大学信息工程学院，北京100876；2．华北电力大学电子与通信工程系，保定071003) 摘要：以公司人事变动领域为例，针对实体关系抽取课题，从知识自动获取角度出发，基于B00tstrapping思想提出 了层次知识获取模型，利用内外2层模块相互嵌套自动获取知识，获得了实体关系分析所需要的领域专用词典和 模板规则．结合全信息理论，对模板添加语义和语用标注，生成全信息知识库．在此基础上，完成关系抽取实验和 评测． 关键词：全信息理论；全信息知识库；层次知识获取；标量聚类 中图分类号：TP391 文献标识码：A ResearchofRelationPatterninSpecificDomain ZHANGSu-xiang1,2, LILei1, TANYong-mei1 (1.SchoolofInformationEngineering,BeijingUniversityofPostsandTelecommunications,Beijing100876,China; 2.DepartmentofElectronicandCommunicationEngineering,NorthChinaElectricPowerUniversity,Baoding071003,China) Abstract:Anewmethodofautomaticentityrelationextractionisproposed.BasedontheBootstrap- pingalgorithm,thehierarchyknowledgeextractionmodelcanbedesigned.Theinnerspecificword extractionmodelandouterpatternextractionmodelcanbenestedeachothertoextractautomatically knowledge,sothatthespecificdictionaryandpatternrulesusedfortheentityrelationextractionis achieved.Combinedwith theComprehensive informationtheory,thesemanticandpragmaticinforma- tion canbeadded intotherelation extractionpatternstogeneratethecomprehensiveinformation knowledge-base(CIKB).Both theexperimentsofrelation extractionandtheevaluationhavebeen done. Keywords;comprehensive information theory;comprehensive informationknowledgebase;hierarchy knowledgeextraction;scalarcluster