摘要： 分析结构化文档的表示方法及检索特点，对一种用于结构化文档检索的贝叶斯网络进行研究。讨论该贝叶斯网络的构造方法、概率估计及推理过程。用网络节点表示文档索引术语和结构单元，用弧表示术语和结构单元的隶属关系，根据TF-IDF方法估计各节点的先验概率，当给定一个查询时，通过计算每个结构单元的条件概率得到该结构单元的相关值。实例验证了该贝叶斯网络的有效性。
关键词:
贝叶斯网络,
结构化文档,
信息检索,
先验概率估计
Abstract: This paper analyzes the representation method and characteristics of retrieval, studies a Bayesian network for structured document retrieval. It discusses the construct method and probability estimates of Bayesian network, and the retrieval process. Index terms and structural units can be represented by nodes, relationship among nodes can be described by arcs, and prior probability of nodes can be estimated according to TF-IDF. The relevance of each structural unit can be computed and ranked by its conditional probability when a user query is given. Example proves the effectiveness of the Bayesian network.
Key words:
Bayesian network,
structured document,
information retrieval,
prior probability estimate
中图分类号:
徐建民, 陈振亚. 一种用于结构化文档检索的贝叶斯网络[J]. 计算机工程, 2011, 37(13): 43-45.
XU Jian-Min, CHEN Zhen-E. Bayesian Network for Structured Document Retrieval[J]. Computer Engineering, 2011, 37(13): 43-45.