The on-the-spot meeting analysis and archiving system

Most meeting analysis and archiving systems are offline-processing systems, in which features and events during meeting are extracted and detected according to the predefined procedure. Owing to the absence of contextual information online guidance in dynamic scene, some important information is missed, which cannot provide active service according to the current situation. In this paper, we develop the on-the-spot meeting analysis and archiving system, which can provide structuring, browsing and retrieving of an archive of automatically analyzed meetings. In the on-the-spot archiving system, structured metadata at different levels are established by extracting the features from the raw data and detecting and analyzing semantic events on the spot, which is based on the framework combining top-down context guidance with bottom-up information processing and multi-cues inference. Compared with current meeting archiving system, our system makes use of the different levels of context and adapt their behavior according to the current situation.