Systematic Verb Stem Generation For Arabic

Performing root-based searching, concordancing, and grammar checking in Arabic requires an efficient method for matching stems with roots and vice versa. Such mapping is complicated by the hundreds of manifestations of the same root. An algorithm based on the generation method used by native speakers is proposed here to provide a mapping from roots to stems. Verb roots are classified by the types of their radicals and the stems they generate. Roots are moulded with morphosemantic and morphosyntactic patterns to generate stems modified for tense, voice, and mode, and affixed for different subject number, gender, and person. The surface forms of applicable morphophonemic transformations are then derived using finite state machines. This paper defines what is meant by ‘stem’, describes a stem generation engine that the authors developed, and outlines how a generated stem database is compiled for all Arabic verbs.