Arabic Stemming techniques as feature extraction applied in Arabic text classification

Abstract

In this paper, we conduct a comparative study about the impact of stemming algorithms, as feature extraction systems, on the task of classification of Arabic text documents. Stemming is forceful and fierce as in reducing words to their three-letters roots. Which may influence the semantics, as various words with divers implications may share the same root. Light stemming, by examination, expels oftentimes utilized prefixes and suffixes in Arabic words. Light stemming doesn’t extract the root and thus doesn’t influence the semantics of words. However, the result of the light stemming is not necessarily a word. For the evaluation, we used corpus contains 5,070 records that fall into six classes. A several tests were done utilizing two separate illustrations of the same corpus. The K-Nearest Neighbors (KNN) classifier was utilized for the classification task. The recall measure is used to evaluate the performance of these methods.