面向军事装备实体的属性抽取  

Attribute extraction for military equipment entity

作　　者：康睿智[1] 郝文宁[1] 程恺[1] 占栋辉[1] Kang Ruizhi Hao Wenning Cheng Kai Zhan Donghui(Center of Simulation ＆ Data, Institute of Command ＆ Information System, PLA University of Science ＆ Technology, Nanjing 210007, China)

机构地区：[1]解放军理工大学指挥信息系统学院仿真与数据中心,南京210007

出　　处：《计算机应用研究》2016年第12期3721-3724,共4页Application Research of Computers

基　　金：江苏省自然科学基金资助项目(BK20150720)

摘　　要：军事装备属性抽取作为构建军用知识库的重要组成部分,具有重要的研究价值和理论意义。针对自由文本中军事装备属性抽取问题,提出了一种面向自由文本的无监督军事装备属性三元组抽取方法。首先利用统计分析的方法对属性三元组在子句中的分布规律进行研究;然后针对分布规律提出一种基于频繁模式挖掘的属性指示词抽取算法,并以抽取出的属性指示词为触发词,结合抽取规则和过滤规则进一步完成三元组的抽取。实验结果表明该方法能够有效地抽取出军事装备属性三元组,其平均准确率达到了88.1%。Attribute extraction of military equipment is an important part of the military knowledge base, which has important research value and theoretical consequences. This paper proposed a new method about unsupervised attribute extraction of mili- tary equipment in free text. Firstly,it analyzed the distribution of the three tuples, and then,proposed a new algorithm of attri- bute indicator word extraction based on frequent pattern mining. It regarded the attribute indicator word as a trigger, combining the extraction rules and the filtering rules, finally made the extraction of three tuples possible. Experiment results show that the algorithm has a good performance and the average accuracy rate reaches 88.1%.

关 键 词：军事装备实体 属性抽取 统计分析 频繁模式挖掘 基于规则 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]