We propose in this work to take advantage of the flexibility of graph oriented databases and the intuitive visual representation of social networks to assist experts in exploring, annotating, and validating face detection and tracking in a 15 year archive of news programs.