所属单位：计算机科学与工程学院

发表刊物：Journal of Intelligent & Fuzzy Systems

项目来源：自选课题

关键字：Entity-Relation Extraction， Multiple Head Selection，Span-Extraction

摘要：In the entity extraction task,there are some complex extraction problems,such as nested entity,entity boundary recognition,context ambiguity,and multi-instanceentity recognition.Entity nesting is an importantchallengein relational extraction.The mainreason of entity nesting problemis that the boundary information betweenen titiesis not clear.Inorder tosolvetheentitynestingproblematthefragmentlevel,whilepreservingtherelationshipbetweenfragmentswiththesame characteristicsandimprovingefficiency,weproposedabrandnewfragmentannotationmethod.Onthebasisoftraditional fragmentannotationmethod,combinedwithpointerannotationmethod,wedesignedanannotationmethodof"ergodic enumeration+groupmapping".Onthebasisofthismethod,anentityextractionmodelisdesigned:Span-Extraction Based Entity Extraction Model(LMA).Our model underwentaseriesofvalidationsin the English datasets NewYork Times(NYT)andWEBNLG,showing signifi can timprovements over the baseline model F1.It can effectively all eviate the above problems.

合写作者：索佳峰,韩东辰

第一作者：赵辉

论文类型：期刊论文

页面范围：1

ISSN号：1064-1246

是否译文：否

发表时间：2023-11-10