I am reseaching the automatic application of semantic tags to video keyframes in unsturtrued video. I am curently working on feature selection tecniques that apply to the unstructed video domain.