This paper presents a comparison of three different speaker recognition methods deployed in a broadcast news processing system. We focus on how the generative and discriminative nature of these methods affects the speaker ...