基于依存分析的离合触发词合法分离形式判定  

Decision on legal separation form of separable triggers based on dependency parsing

作　　者：肖升[1,2] 李勇帆[2] 何炎祥[1] 

机构地区：[1]武汉大学计算机学院,武汉430079 [2]湖南第一师范学院信息科学与工程系,长沙410205

出　　处：《计算机工程与应用》2014年第10期11-17,共7页Computer Engineering and Applications

基　　金：教育部人文社会科学规划基金项目(No.12YJAZH065);湖南省高校科技创新团队支持计划资助项目(湘教通[2010]212号)

摘　　要：离合触发词的构词语素可能因插入、颠倒、省略而产生多种合法分离形式,这些分离形式与原形一样也能表征事件。为完整抽取事件,提出一种基于依存分析的离合触发词合法分离形式判定算法。该方法首先借助依存分析考察离合触发词合法分离形式在句中所受的依存约束,然后将这些约束转化为可计算的判定规则,最后利用判定规则对离合触发词的合法分离形式进行判定。实验结果显示,排除稀疏数据前,此方法的正确率、召回率、F值分别为82.2%、88.3%、85.1%;排除稀疏数据后,正确率、召回率、F值提升到82.4%、88.7%、85.4%。方法已基本具备应用潜质。The morphemes of separable triggers may produce a variety of legal separation form due to the insertion, invertion, ellipsis, and the legal separation form can also characterize events as the primary form of separable triggers. A decision algorithm of legal separation form of separable triggers based on dependency parsing is proposed, for the purpose of complete extraction of events. Firstly some dependency constraints of the legal separation form of separable triggers are found with the help of dependency parsing. Then some computable decision rules are extracted from those dependency constraints. Finally decision on the legal separation form of separable triggers are executed using decision rules. Experimental results show that the precision, recall, F-Measure of this approach is 82.2%, 88.3%, 85.1%;and the precision, recall, F-Measure of this approach is 82.4%, 88.7%, 85.4%exclusion of sparse data, so the approach has the application potential basically.

关 键 词：离合词 离合触发词 分离形式 依存分析 事件抽取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]