Keywords and Keyphrases are very important to capture the semantics contained in texts. Their extraction is a topic of particular relevance to a great number of researchers. Keywords and Keyphrases are useful for many applications, like: Indexation, Information Retrieval, Classification, Recommendation, Contextual advertising, etc. Their extraction is a challenging problem. Many approaches have ... [Show full abstract] been proposed for Keywords/Keyphrases extraction for several languages, including the Arabic language. However, the question remains to: what is the most performing approach in this field of research?. In this paper, a comparison is performed between two different methods applied to Keyphrase extraction from Arabic Texts: namely, Naïve Bayes and an approach based on a graph model. The experiments and the results obtained attest that the Graph based approach is more performing than Naïve Bayes. For the first approach the F-score is close to 70%; while the second one obtains an F-score of less than 64%.