Audio and video news alert system
Growing numbers of media outlets and channels have contributed to the phenomenon of information overload, leading to a subsequent demand for customised filtering of the vast quantity of news. Furthermore, there are indications of a growing competition to be 'the first to know', now a necessity in many professions. News alert systems tend to lack content from audio and video sources. The ALERT project, funded under the EU's IST programme, has associated state-of-the-art speech recognition with audio and video segmentation, as well as automatic topic indexing. Based on these technologies the project has developed a demonstration version of an automatic media-monitoring system in three languages: French, German and Portuguese. The system includes a topic-specific news alert system, which can be tuned to function according to individual user preferences. This news alert system continuously monitors TV channels, and searches inside their news programmes for stories that match the profile of a given user. The system can automatically detect the start and end of a broadcast news programme. When the programme is detected, the system automatically records and segments it into individual shots. The speech recognition system then transcribes the audio track to produce a text version of the programme's narration. The ALERT system goes on to index, summarise and store the programme's video, audio and related text. Following content processing, the system then searches in the user profile database for profiles, set up in advance by users, which correspond to the topics of indexed programme content. If any topic matches the user preferences for subjects of interest, an email is send to that user indicating the occurrence and location of one or more stories on the selected topics. This alert message enables a user to follow the links to the relevant video clips. The new technology is likely to impact news alert services and the way users access news.
