一种改进的元搜索排序合成算法

An Improved Rank Merging Algorithm for Meta Search

为了提高元搜索引擎的查询精度,提出了一种改进的元搜索结果合成算法.首先,通过分析搜索结果列表中包含的文本信息,综合考虑搜索结果与查询的匹配完全程度和相关程度,给出了文本分析的规范化方法;并结合搜索结果的排序信息计算文档的相关分值,据此实现对局部相似度的调整.然后,利用成员搜索引擎的性能评价,提出了改进的影子文档方法来估算非相关文档的相关分值.最后,采用基于群决策的合成方法对搜索结果进行一致性排序.实际Web环境中的测试表明,所提出的算法比现有合成算法具有更好的搜索结果相关性.

In order to improve the precision of meta search engine, an improved merging algorithm of meta search results is proposed. In this algorithm, first, the text-based information obtained from search results is analyzed and both the query-match grade and the result relevancy are considered to give an approach on the text normalization for meta search. Next, the relevant scores of documents are normalized by incorporating text analysis with the ranks given by the search engines for the purpose of adjusting the local similarities. Then, based on the performance evaluation of underlying search engines, an improved shadow document method is presented to evaluate the scores of non-relevant documents. Finally, a merging method based on the group decision making is adopted to sort the search results. It is found from the tested results in an actual Web environment that the search results obtained by the proposed algorithm are of higher relativity than those by the existing merging algorithms.