Connecting word clusters to represent concepts with application to web searching

Abstract

The need to have a search technique which will help a computer in understanding the user, and his requirements, has long been felt. This paper proposes a new technique, of doing so. It proceeds by first clustering the English language words into clusters of similar meaning, and then connecting those clusters according to their observed relationships and co-occurrences in web pages. These known relationships between word clusters are used to enhance the user's query, and in effect 'understand' it. This process will result in giving results of more value to the user. This procedure does not suffer with the problems faced by many of the presently used techniques.