Show simple item record

Author Duwairi, Rehaben_US
Author Al-Refai, Mohammaden_US
Author Khasawneh, Natheeren_US
Available date 2009-12-28T05:33:38Zen_US
Publication Date 2007-11-18en_US
Publication Name 4th International Conference onInnovations in Information Technology 2007
Citation Duwairi, R.; Al-Refai, M.; Khasawneh, N., "Stemming Versus Light Stemming as Feature Selection Techniques for Arabic Text Categorization," Innovations in Information Technology, 2007. IIT '07. 4th International Conference on , vol., no., pp.446-450, 18-20 Nov. 2007en_US
URI http://dx.doi.org/10.1109/IIT.2007.4430403en_US
URI http://hdl.handle.net/10576/10501en_US
Abstract This paper compares and contrasts two feature selection techniques when applied to Arabic corpus; in particular; stemming, and light stemming were employed. With stemming, words are reduced to their stems. With light stemming, words are reduced to their light stems. Stemming is aggressive in the sense that it reduces words to their 3-letters roots. This affects the semantics as several words with different meanings might have the same root. Light stemming, by comparison, removes frequently used prefixes and suffixes in Arabic words. Light stemming doesn't produce the root and therefore doesn't affect the semantics of words; it maps several words, which have the same meaning to a common syntactical form. The effectiveness of above two feature selection techniques was assessed in a text categorization exercise for Arabic corpus. This corpus consists of 15000 documents that fall into three categories. The K-nearest neighbors (KNN) classifier was used in this work. Several experiments were carried out using two different representations of the same corpus; the first version uses stem- vectors; and the second uses light stem-vectors as representatives of documents. These two representations were assessed in terms of size, time and accuracy. The light stem representation was superior in terms of classifier accuracy when compared with stemming.en_US
Language enen_US
Publisher IEEEen
Subject Arabic languageen_US
Subject K-nearest neighbors classifieren_US
Subject feature selectionen_US
Subject light-stemmingen_US
Subject stemmingen_US
Subject text categorizationen_US
Title Stemming Versus Light Stemming as Feature Selection Techniques for Arabic Text Categorizationen_US
Type Conference Paperen_US


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record