基于深度学习和公开来源信息的反恐情报挖掘  

Anti-terrorism Intelligence Mining Based on Deep Learning and Open Source Information

作　　者：郭璇[1,2] 吴文辉[2] 肖治庭[2] 袁宏国 

机构地区：[1]武警警官学院,四川成都610213 [2]国防科技大学信息通信学院,湖北武汉430010

出　　处：《情报理论与实践》2017年第9期135-139,共5页Information Studies:Theory & Application

基　　金：国家社会科学基金西部项目"基于信息网络的反恐预警体系架构及关键问题研究"(项目编号:16XSH003);国家社会科学基金军事学项目"武警部队基于信息网络的反恐预警研究"(项目编号:16GJ003-245)的阶段性成果

摘　　要：[目的/意义]反恐情报工作通常依赖于情报人员的智力和经验,但大数据时代对海量数据的处理仅仅依靠人力显然是不够的。为了提高反恐情报信息工作效率,从而全面提升反恐情报预警能力,迫切需要自动化智能化的情报处理技术来减轻情报人员的工作量。[方法/过程]利用人工智能领域最新成果深度学习技术对公开来源信息进行反恐情报挖掘和分类,并采用深度学习词向量工具Word2vec对大量开源中文语料库进行深度训练。[结果/结论]实验结果证明深度学习算法在反恐情报的挖掘分类精度上要优于传统情报分类算法;通过Word2vec计算相似词向量的余弦距离,利用已知的恐怖分子、恐怖组织名单来挖掘与其密切关联的潜在恐怖分子,对于反恐情报预警工作具有积极意义。[局限]深度学习算法需要大量样本进行训练,时间开销稍大。[ Parpose/significance ] Anti-terrorism intelligence work usually depends on the intelligence and experience of intel- ligence personnel, but massive data processing in the era of big data demands far more than manual work. In order to improve the work efficiency of anti-terrorism intelligence work and comprehensively enhance the ability of anti-terrorism intelligence warning, it is imperative to adopt automotive and intelligent information processing technology to reduce the workload of intelligence personnel. [ Method/process ] This paper uses the deep learning technology to mine and classify the open source information, and adopts deep learning tool Word2vec to train a large number of open source Chinese corpora. [ Result/conclusion] Experimental results show that the deep learning algorithm is superior to the traditional machine learning algorithm in the classification accuracy of anti-terrorism information mining. Through the calculation of the cosine distance of similar words vectors by Word2vec, the potential terrorists closely associated with the known terrorists and terrorism organizations are discovered, which has positive significance for anti-terrorism intelligence early warning work. [ Limitations ] The deep learning algorithm requires a large number of samples for training, so the time cost is larger.

关 键 词：反恐情报 预警分析 词向量 信息挖掘 

分 类 号：G350[文化科学—情报学]