多元表征军事信息可信度研究  

Research of the Reliability on Multi-Representations Military Information

作　　者：张天宇[1] 林鸿飞[1] 

机构地区：[1]大连理工大学计算机科学与技术学院,辽宁大连116024

出　　处：《计算机工程与科学》2011年第9期109-116,共8页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60673039;60973068);国家863计划资助项目(2006AA01Z151);教育部留学回国人员科研启动基金;高等学校博士学科点专项科研基金资助课题(20090041110002)

摘　　要：伴随互联网的迅猛发展,网页成为信息主要承载形式,各国的军事机构开始对网页信息中的军事情报进行搜集和监管,如何快速准确地从互联网获取所需军事信息成为研究的热点。本文针对互联网军事信息数据量大、更新速度快且真伪难辨的特点,提出了多元表征军事信息可信度的方法。首先,在对军事信息可信度表征研究的基础上,提出了基于信息源、信息评论和信息内容军事信息可信度表征方法;而后,引入排序学习的思想,将表征结果作为特征对独立表征结果进行了融合,获得较优的信息可信度排序结果。最终,通过对比实验证明,本文提出的三种表征方法均有益于军事信息的获取和甄别,能很好地满足军事研究人员的需求,且多元表征融合比独立表征效果更好。With the rapid growth of the Internet, web page becomes the main form of information presentation Many nations military institutions pay more attention to searching and overseeing the military messages on web pages How to gain the military information on the Internet quickly and accurate- ly becomes more important than ever. This paper proposes a multi-representation reliability method to resolve the problem. First based on the research on the military representation reliability, this paper proposes a method which focuses on information resources, reviews and contents. Second we fuse inde- pendent representative results by regarding them as features according to sort learning~ And then we get a better performance on reliability sorting. Finally, the experimental results show that the method this paper proposed performs military information acquiring and discriminating well to satisfy the demands of military researchers, and the multi-representation method performs better than the independent-repre- sentation methods.

关 键 词：多元表征 军事信息 可信度 排序学习 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]