Application of natural language processing and speech processing technology to production of closed-caption TV programs for the hearing impaired

Television service is indispensable to a human life in the modern age. The people who are seeing or hearing impaired, however, do not enjoy the TV programs as much as they want. In closed-caption service, the speech in TV programs are put into characters and superimposed on TV pictures for the bene t of hearing impaired people. We proposed in the paper natural language and speech processing techniques by which the efcient production of the TV programs with closed-captions can be available. By our method, an electronically available text of TV news programs will be summarized and synchronized with the speech and video automatically, then superimposed on the original programs. For the summarization, we propose a shallow method which detects the important words in the news text and determine which section of the text is important. We also propose the keyword pair model for the synchronization between speech and text. We described the outline of our project and the results of preliminary experiments.