基于文档关系的扩展信念网络检索模型  

An Extended Belief Network Retrieval Model Based on Document Relationships

作　　者：徐建民 何丹丹 吴树芳[2] Xu Jianmin;He Dandan;Wu Shufang(College of Cyberspace Security and Computer,Hebei University,Baoding 071002;School of Management,Hebei University,Baoding 071002)

机构地区：[1]河北大学网络空间安全与计算机学院,保定071002 [2]河北大学管理学院,保定071002

出　　处：《情报学报》2019年第11期1160-1165,共6页Journal of the China Society for Scientific and Technical Information

基　　金：国家社会科学基金后期资助项目“基于术语关系的贝叶斯网络检索模型扩展”（17FTQ002）;河北省自然科学基金项目“基于贝叶斯网络的话题识别与追踪方法研究”（F2015201142）

摘　　要：合理利用文档关系可以提高模型的检索性能。针对基本信念网络检索模型未考虑文档关系的不足,通过在基本模型上增加一层文档节点,提出一种具有两层文档节点的扩展信念网络检索模型,给出了模型的拓扑结构和概率推导。在拓扑结构中,术语与查询的关系、术语与文档的关系和两层文档之间的关系都用弧来表示,其中文档关系依据文档相似度确定。在概率推导中,利用文档相似度及文档节点的父文档个数对原模型的概率推导做出修正,使得检索概率更为准确。实验采用折损累积增益值和查准率-查全率曲线来评价扩展模型的性能,结果表明,扩展模型使得相关文档排名更合理,并且在保证查全率的条件下提高了查准率。The performance of a retrieval model can be improved by using relationships among documents reasonably. To solve the problem of a basic retrieval model that does not use document relationships, an extended belief network retrieval model with two layers of document nodes is proposed. The topology structure and probability inference of the extended model are given. In the topology structure, relationships between items like terms and queries, terms and documents, and two layers of documents, are indicated by arcs. The relationship between documents is determined by the similarity of the documents. In the probability inference, the retrieval probability is made more accurate by using document similarity and the number of parent documents to modify the original probability. In experiments, the value of discounted cumulative gain and the precision-recall curve are introduced to attest to the performance of our proposed extended model. The results show that the extended model makes the ranking of related documents more reasonable and improves the precision under the premise of guaranteeing recall.

关 键 词：信念网络 文档关系 检索模型 文档相似度 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]