文章摘要
李芊芊,张克亮.基于依存分析的军事领域英文实体关系抽取研究[J].情报工程,2019,5(1):098-112
基于依存分析的军事领域英文实体关系抽取研究
Entity Relation Extraction Based on Dependency Parsing in Military Field
  
DOI:10.3772/j.issn.2095-915X.2019.01.009
中文关键词: 关系抽取;依存分析;英文实体关系抽取;军事文本
英文关键词: Relation extraction; dependency parsing; English entity relation extraction
基金项目:国家自然科学基金项目“ 多语种言语识别项目”(11590771)。
作者单位
李芊芊 战略支援部队信息工程大学 
张克亮 战略支援部队信息工程大学 
摘要点击次数: 3195
全文下载次数: 2737
中文摘要:
      在海量信息的分析、处理中,实体关系抽取技术发挥着至关重要的作用。由于英文实体关系抽取存在信息量少和关系表述内聚性低的问题,针对英语军事文本,提出一种改进的基于依存分析的英文实体关系抽取方法。首先根据依存句法分析识别句子中的谓词,然后提取该谓词支配的所有依存关系,接着对上述依存关系进行二级扩展,得到完整的语、宾语、介词短语,同时对状语和补语从句进行二级扩展,得到从句中的谓词并再次进行二级扩展,最后根据三元组的构成模式输出关系三元组。在英语军事领域语料上的实验表明,提出的方法具有较高的准确率和召回率,与斯坦福coreNLP3.6.0 关系抽取方法相比,F 值提高了57.71%。
英文摘要:
      Entity relation extraction plays an important role in the analysis and processing of mass information. To reduce incoherent and uninformative extractions, an improved method of English entity relation extraction based on dependency parsing is proposed. The predicate of the sentence is firstly recognized according to the results of dependency parsing. Then,all dependency relations of the predicate are further extended to output complete subject, object and prepositional phrase.Meanwhile, the adverbial clause and complement clause is further extended to find and extend the predicate in the clause.Finally, the relation triples are output according to the pattern of triples. Experiments on the military corpus show the proposed method has a higher precision rate and recall rate, and the F m easure is 57.71% higher than Stanford coreNLP3.6.0 method.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮