Abstract: Aiming at the problem of manual annotation in TBVR and the immatureness of CBVR and in-depth analysis on the requirements and features of video retrieval in virtual television news, a news video retrieval method based on multiple keywords was proposed. Annotation dictionary, tree annotation structure, automatic keywords extraction, similarity computation model and automatic retrieval algorithm were discussed in detail. An experiment was made on this method. The result shows that it has attained a very high recall rate with a comparative high accuracy rate and it can solve the problem of automatic video retrieval in virtual television news.

Key words: Virtual television news, Video retrieval, Video similarity, Strategic seminar gaming