No abstract

No abstract

Conventional information retrieval systems have proven ineffective in dealing with information overload. One possible solution is to incorporate some features that allow users of these systems to custom handle this information. In order to enable systems of this kind, some of the characteristics of present-day systems should be reviewed. Among other features, all documents are described with the same level of detail. We believe that the redrafting of document models is the starting point for reform of these systems. The paradigm of granular computing has proven to be very suitable for the treatment of complex problems and can produce significant results in large-scale environments such as the Web. This paper explores the granulation process of words with a view to its application in the subsequent improvement in document representation. We use fuzzy relations and spectral clustering in this process and present some result.