back

w3c

Fifth International World Wide Web Conference

Precise and Informative Clustering

More precise:
Better measure of documents' profile similarity by using a less ambiguous indexing unit: the lexical affinity (LA)
More informative:
Associate to each cluster formed a list of key concepts and a pseudo-title derived from the LA-based profiles

single word	norm. freq	lexical affinity	norm. freq
book	0.399	book softcopy	0.346
softcopy	0.394	cd rom	0.336
ibm	0.343	collection kit	0.263
library	0.218	ibm softcopy	0.217
cd	0.161	ibm library	0.206
rom	0.154	book library	0.201
bookmanager	0.147	ibm program	0.191
pc	0.142	library softcopy	0.181
bookshelf	0.129	bookmanager ibm	0.170
collection	0.129	book bookshelf	0.165
program	0.127	kit softcopy	0.139
kit	0.127	host system	0.139
information	0.115	collection softcopy	0.134
system	0.100	library program	0.108
host	0.093	library reader	0.103
product	0.090	book print	0.103
picture	0.080	cd softcopy	0.098

Single word profile vs. LA profile for "IBM softocopy library package"

Automatically Organizing Bookmarks per Contents (Y.S.Maarek,I.Z.Ben-Shaul)