A new stemming algorithm dedicated for Arabic documents Classification

Abstract

stemming is the process of mapping several words related semantically but different morphologically to one word root. Therefore it is very important preprocessing step before perform any natural language processing task such as text classification for reducing and getting accurate features. Arabic language possess spatial morphological characteristics make stemming process difficult. Therefore many light and root stemmers proposed for handling this problem but unfortunately most of them appeared confusion between the original letters and affixes may attached with root of the words. This paper introduced A new stemming algorithm dedicated for Arabic documents classification (NSAAD) it is dedicated for stemming Arabic words using predefine patterns and affixes lists. Experiments performed on lists of words related in their root but different in morphological form. Comparing with Information Science Research Institute's (ISRI) stemmer, proposed technique introduced better results.