… To address the drawbacks of the previous work on Arabic text processing, I propose an
alternative methodology that defines rules for stemming words instead of chopping off the
letters. The rules are automatically learned by clustering morphologically related words and
relying on the input documents only as a content based dictionary. I introduce the “local stem”
as a replacement to the current stemming representations and use the concept as a both a
stemming algorithm and a query expansion tool. Finally, this research is directed at …