Comparative Analysis and Evaluation of Stemming and Preprocessing Techniques for Arabic Text

Abstract

Arabic information retrieval is challenging due to the language's complex morphology and syntax. Preprocessing and stemming improve the accuracy and efficiency of Arabic information retrieval. This paper provides a comprehensive analysis of the existing literature on Arabic preprocessing and stemming techniques. The paper identifies the limitations and challenges of these techniques. The paper emphasizes the importance of preprocessing and stemming and underscores the need for further research to improve Arabic information retrieval. This study evaluates ten stemmers on a public dataset. The results show that root-based stemmers: Lucene, and khoja got the highest reduction rate 90.9%, and 85% respectively. The results emphasize that root-based stemmers have good conflating ability for similar terms, while light-based stemmers under-stem similar terms.