面向军事装备实体的属性抽取

发布时间：2018-04-10 05:03

  本文选题：军事装备实体　切入点：属性抽取　出处：《计算机应用研究》2016年12期

【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%。
[Abstract]:As an important part of constructing military knowledge base, attribute extraction of military equipment has important research value and theoretical significance.In order to solve the problem of attribute extraction of military equipment in free text, a method of unsupervised attribute extraction of military equipment for free text is proposed.Firstly, the distribution of attribute triples in clauses is studied by means of statistical analysis, and then an algorithm for extracting attribute deixis based on frequent pattern mining is proposed.Taking the extracted attribute indicator as the trigger word and combining the extraction rule with the filtering rule, the triple extraction is further completed.The experimental results show that the proposed method can extract the attribute triples of military equipment effectively, and the average accuracy is up to 88.1g.
【作者单位】： 解放军理工大学指挥信息系统学院仿真与数据中心;
【基金】：江苏省自然科学基金资助项目(BK20150720)
【分类号】：TP391.1
，

本文编号：1729754