Introduction to Clustering Large and High-Dimensional Data
Focusing on a few of the most important clustering algorithms, this book provides a detailed account of these major models in an information retrieval context.
TABLE OF CONTENTS Introduction to Clustering Large and High-Dimensional Data
Chapter 1: Introduction and Motivation
Overview
For a given document collection 
= { D 1 , D 2 , ..., D m} and a query Q one often is concerned with the following basic problems:
-
Find documents in
related to the query. If, for example, a distance between two documents D i and D j is given by the function d( D i , D j) and a threshold tol > 0 is specified one may be interested in identifying the document subset tol ? defined by
-
Partition the collection
into disjoint subcollections ? 1 , ? 2 , ..., ? k (called clusters) so that the documents in a cluster are more similar to each other than to documents in other clusters. The number of clusters k also has to be determined.
When tight clusters ? i, i = 1 , ..., k are available representatives c i of the clusters can be used instead of documents to identify tol. The substitution of documents by representatives reduces the data set size and speeds up the search at the expense of accuracy. The tighter the clusters are the less accuracy is expected to be lost. Building high quality clusters is, therefore, of paramount importance to the first problem. Applications of clustering to IR are in particular motivated by the Cluster Hypothesis which states that closely associated documents tend to be related to the same requests.
Copyright Jacob Kogan 2007 under license agreement with Books24x7
