Analysis of the Complementarity of Latent and Concept Spaces for Cross-Modal Video Search