基于Markov网络的结果重排技术

信息检索中通过网页链接信息提取文档内部关系进行搜索结果重排可以提升检索系统的性能。通过Markov网络来展现文档内部关系，该网络更直观地解释了文档间的语义相关性，利用这种文档内部语义关系计算文档重要性对检索结果进行重排。根据文档分布特征阐述了Markov文档网络的构造算法，讨论了Top-k及其相关文档的重要性评分算法，修正初始检索的文档评分。通过这种方式，既保持了文档图的查询相关性，又丰富了文档内部关系，扩大了重排序范围。实验表明，在多个标准文档集上基于Markov网络的结果重排技术对检索性能有较大的稳定提升。

Document re-ranking based on Markov network

In information retrieval, search results re-ranking can improve retrieval system through internal document relation extracted from webpage link information. In this paper we propose a document re-ranking technique based on Markov network, which can explain the sematic relationship between documents better, and can be used to compute the document importance, which is combined with the initial search results into re-rank stage. First, we analyze the algorithm of constructing document Markov network, and then explain the scoring method of Top-k and its related documents, to revise the initial retrieval score. The experiment results on several different data sets demonstrate the effectiveness of the proposed document re-ranking method. Compared to the traditional PageRank method, it has stable advantage.