Next, the system constructs a set of components to characterize the document, wherein the set of components includes components for candidate clusters.

• 然后，所述系统构造一组分量以描述所述文档，其中该组分量包括候选聚类的分量。