Machine learning of event segmentation for news on demand
Communications of the ACM
| Hi-index | 0.01 |
This poster describes methods to enable intelligent access to multimodal information streams. We illustrate these methods in two integrated systems: the Broadcast News Editor (BNE) which incorporates image, speech, and language processing and the Broadcast News Navigator (BNN) which provides search, visualization and personalized access to broadcast news video. BNN enables users to perform keyword and named entity search, temporally and geospatially visualize entities and stories, cluster stories, discover entity relations, and obtain personalized multimedia summaries. By transforming access from sequential to direct search and providing hierarchical hyperlinked summaries, BNE and BNN enable users to access topics and entity news clusters nearly three times as fast as direct search of video.