A novel approach for indexing Arabic documents through GPU computing

Abstract

In contrast to English search engines, Arabic search engines did not have their fair share in modern studies despite the continuous growth of Arabic Internet users and data. Towards bridging the gap, this paper presents a novel indexing algorithm customized for Arabic documents. Our algorithm exploits the characteristics of the Arabic language to enhance indexing and lookup. Additionally, the algorithm utilizes the highly parallel architecture of the graphics processing unit to speed-up the indexing. Finally, we discuss some of the synchronization challenges we faced and the techniques we used to overcome them. The preliminary tests of our GPU-accelerated Arabic indexer show promising speed-up factors. © 2012 IEEE.