摘要：In this paper, we propose an Arabic Human-machine oral dialogue manager. This work enters in the context of SARF system ([4], [9]), an interactive vocal server that provides information on Tunisian railway using modern standard Arabic. The proposed dialogue manager is based on a structural approach and uses a formal grammar that we transformed into a deterministic finite automaton to monitor and maintain the dialogue with the user. The evaluation results of this dialogue manager are very encouraging. Thus, the success rates of tasks and of the dialogue act identification are respectively 89.65% and 98.31%. The performance average time is equal to 0.0779 seconds.
