This paper is dedicated to the study of automatic acoustic segmentation systems, often required for broadcast news show processing. This study is carried in the framework of the phase I of the French evaluation campaign, called ESTER. It aims at evaluating the behavior of an acoustic segmentation system, especially tuned of American broadcast news show processing, when it is applied to French ones. Furthermore, different implementations of this system are evaluated, each of them taking into account French data. The goal of these implementations is to demonstrate which kind of French data have to be used and how to integrate them in the acoustic segmentation system. Results obtained through various experiments highlight interesting issues.