Arabic Stemming Without A Root Dictionary.pdf

Arabic Stemming Without A Root Dictionary Kazem Taghva, Rania Elkhoury, Jeffrey Coombs Information Science Research Institute University of Nevada, Las Vegas Las Vegas, NV 89154-4021 taghva@ Abstract We have implemented a root-extraction stemmer for Ara- bic which is similar to the Khoja stemmer but without a root dictionary. Our stemmer was found to perform equivalently to the Khoja stemmer as well as so-called “light” stemmers in monolingual document retrieval tasks performed on the Arabic Trec-2001 collection. A root dictionary, therefore, does not improve Arabic monolingual document retrieval. 1 Introduction The two most successful approaches to Arabic stemming have been a root-extraction stemmer developed by Khoja[9] and a “light” affix removing stemmer developed by Larkey et al.[10, 11]. Larkey et al. have shown that the Khoja and Light stemmers, aswell as co-occurrence analysis-aided stemmers, perform information retrieval tasks with statisti- cally equivalent precision.[10] We show that a root-extraction stemmer similar to Khoja’s can be constructed without some of the short- comings of that stemmer. Most obviously our stemmer does not require a root dictionary which can be difficult to maintain. Our stemmer was shown to be just as effective as the Khoja stemmer for information retrieval tasks. Section 2 presents basic concepts concerning stemming in general and the Arabic stemming problem in particular. Previous work on Arabic Stemming is discussed in section 3 along with a brief presentation of the algorthm used in the Khoja stemmer. Our Arabic stemmer is presented in sec- tion 4. (The complete algorithm is listed in Appendix A.) We describe and evaluate our experiments in section 5. Fi- nally, we present our conclusions and prospects for future work in section 6. 2 Background A stemmer for the context of this paper is an automatic process in which morphological variants of terms are mapped to a single representative string called a stem. More specifically