Indexing and retrieval of TV broadcast news using DANCERS

In this article we present DANCERS, a system for automated indexing and ret rieval of TV broadcast news. The system output, a topic-based report organi zation structure, is obtained in three major steps. In the first step, the visual and the speech streams of a news program are analyzed in order to pa rtition the program into report segments. In the second step, the transcrib ed speech of each of the report segments is matched with the content of a l arge prespecified textual topic database. This database covers a large numb er of topics and can be updated anytime, e.g., according to the user's inte rests. The result of the matching procedure is a list of most probable topi cs per report segment that are used in the last step to merge related neigh boring segments into reports. The most probable topics that serve as the ba se for the segment-merging procedure are at the same time the indexes of th e newly created reports, using which the reports can be retrieved or classi fied A relative simplicity of the indexing approach, minimized user involve ment and highly satisfying indexing performance are the major characteristi cs of our system. (C) 2001 SPIE and IS&T.