An interval-based method for text clustering

Abstract

This paper describes a clustering approach for classifying texts where dynamic text are evaluated with intervals (interval-valued numbers or IVN) rather than numbers. This new interval measurement allows us to capture the text relevance by multiple keywords at the same time and thus can provide multi-dimension text ranking. We build several comparison mechanisms to rank the intervals based on interval multiple values, their distributions and relationships. Then, the dynamic texts in information systems can be pro-actively clustered into collections usingthis ranking results. © Springer Science+Business Media B.V. 2010.