Methods and apparatuses of the present invention generally relate to generating actionable data based on multimodal data from unsynchronized data sources. In an exemplary embodiment, the method comprises receiving multimodal data from one or more unsynchronized data sources, extracting concepts from the multimodal data, the concepts comprising at least one of objects, actions, scenes and emotions, indexing the concepts for searchability; and generating actionable data based on the concepts.