A Document as a Small World

The small world topology is known widespread in biological, social and man-made systems. This paper shows that the small world structure also exists in documents, such as papers. A document is represented by a network; the nodes represent terms, and the edges represent the co-occurrence of terms. This network is shown to have the characteristics of being small world, i.e., highly clustered and short path length. Based on the topology, we develop an indexing system called KeyWorld, which extract important terms by measuring their contribution to the graph being small world.

• Type: Conference
• Year: 2001
• Where: JSAI
• Authors: Yutaka Matsuo, Yukio Ohsawa, Mitsuru Ishizuka
• Keywords: Artificial Intelligence, JSAI 2001, Small World, Small World Structure, Small World Topology