In the multidimensional space of documents that are represented by vectors, clustering can be viewed as partitioning. When the clustering algorithm allows overlap, however, boundaries are not as sharp as illustrated since a document can be in several clusters.