ACMMM 2015: Deep Multimodal Speaker Naming