130739Building on subtitles as a source of live metadata for UK broadcast TVClosedFast TrackInnovate UKTV Subtitles are already a rich source of pre-existing metadata. This has been exploited by some to provide advanced search techniques for video archives such ashttp://www.hulu.com/labs/captions-search. Modern Natural Language Processing (NLP) techniques make it possible to extract more detailed and structured information from unstructured scripts. This includes indentifying people, places, political concepts etc. We propose to create a system that will extract subtitles from the broadcast stream and feed them through a concept extraction engine in real time to create a new layer of semantic metadata