Arabic root extraction through generation and filtering

In this paper, we propose a new approach to extract the roots of Arabic words. Our approach is based on two phase process: (1) generation phase: where an initial set of candidate roots is generated and (2) filtering phase: where the candidate root set that was generated at the first phase is filtered to eliminate incorrect roots. The main advantages of this approach is its ease to implement and to expand to cope with new derivation forms (patterns). New filters can easily be added to eliminate roots that are known to be incorrect.