Contextual Information Extraction for Video Data