ON CLUSTERING TECHNIQUES OF CITATION GRAPHS
Abstract
The study is an application of graph theory to the problem of clustering in document retrieval systems using bibliographic coupling devices. The problem is attacked by mapping the citation graph of the document collection into a unidimensional storage array. The figure of merit of the location assignment is the total distance between connected pairs of documents, or, equivalently, the 'stretching' resulting from the mapping. This is the objective function of the problem. An algorithm is then presented for the reduction of the objective function, which provides a currently improving solution. Its computational complexity only grows as N3/2, where N is the collection size.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1967
- Accession Number
- AD0652593
Entities
People
- Franco P. Preparata
- Robert Tienwen Chien
Organizations
- University of Illinois Urbana–Champaign