Original language | English |
---|---|
Pages (from-to) | 423-430 |
Journal | Soft Computing |
Volume | 10 |
Issue number | 5 |
DOIs | |
Publication status | Published (in print/issue) - 1 Mar 2006 |
Bibliographical note
Other Details------------------------------------
This paper investigates the strengths of k-nearest neighbour (k-NN) and Rocchio learning algorithms, and develops a new learning method called kNNModel for text categorization, which combines the strengths of KNN with those of Rocchio. A text categorization prototype system was developed within the EU FP5 Intelligent Content Management System (ICONS) project (IST-2001-32429), comprising kNNModel, kNN, Rocchio and Support Vector Machine (SVM). The kNNModel approach provides an effective tool for text indexing, which is an essential component of search engines, and the prototype system is being used as a benchmark system for developing new methods and techniques for text categorization.