Arabic named entities (ANE) are often sources of information. This is why several applications of natural language processing (NLP) mainly in information retrieval. In order to improve the relevance of the information obtained, links to free resources can be established. Indeed, the recognition of these entities requires the use of adequate formalisms. In this paper, we propose an approach based on transducer cascades. This approach allows the recognition of ANE more precisely the dates. This category can be an integral part in the events and the names of places. The implementation of the developed transduc-ers cascades elaborated by using the CasSys tool available under the Unitex platform. The results were encouraging.