Show description

From these tables, it is shown that the descriptive labels of latent factors are characterized by some “prominent” pages whose probabilistic weights are exceeding one predefined threshold. This work is done by interpreting the contents of corresponding pages since these “dominant” pages contribute greatly to the latent factors. With the derived characteristic factor, we may semantically discover usage-based access pattern. Table 1. 3 Examples of Web Page Categories At this stage, we utilize aforementioned clustering algorithm to partition Web pages into various clusters.

Let C be the set of vertices corresponding to the approximate community. Let (X, V \ X) be the min-cut of G corresponding to C, where s ∈ X ⊆ V . Thus C = X \ {s}. For a vertex v, we denote by d+ (v) and d− (v) the out-degree and in-degree of v in G , respectively. We consider a situation in which the following condition (1) hold: (1) G has no edge (u, v) for any pair of vertices u and v such that u, v ∈ P or u, v ∈ Q. We then claim that the following (A) and (B) hold for the approximate community found by the original max-flow based method.

Fig. 3. Examples of element relationship type Element label: In Figure 3, attribute “Publication Year” has two elements whose labels are “after” and “before” respectively. In this case, “Publication Year” is treated as the label of the attribute. Element labels are considered as the child labels of their attribute, which usually represent the semantics of the elements. Logic relationship of attributes: Attributes on a search interface can be logically combined in different ways to form a query to access the underlying database.