一种基于微分流型的Web检索算法

为进一步提高Web的检索性能,提出了一种基于微分流型的 Web检索算法.该算法的核心思想为:用微分流型描述Web空间,用流型的切平面模拟Web的切空间,将流型上2点之间的测地线距离转化为二维平面上点与点之间的欧氏距离,基于距离的匹配过程由原来的在整个文档集合中进行,转换为仅在查询的邻域内进行,以减少匹配过程的计算量.实验结果表明,该算法的查准率比 Google和 IIT分别提高 35 24%和40 93%,它更适合于大型异构Web文档集合的检索.

Web Retrieval Algorithm Based on Differential Manifold

To further improve the retrieval performance of Web retrieval, an algorithm of Web retrieval based on differential manifold was proposed. The essential idea of the algorithm is as follows. Firstly, Web space was modeled as a manifold, and Web tangent space was simulated using a tangent plane of the manifold. Secondly, geodesic distance between points in the manifold was transformed into Euclidean distance between points in the 2-dimension plane. Finally, distance-based matching was performed only in the neighborhood of a query rather than in the whole document collection so as to the computational complexity of the matching process was reduced. Experimental results show that the algorithm surpasses over Google and IIT 35.24% and 40.93% in average precision respectively, and it is more appropriate for large and heterogeneous Web space than other algorithms.