Managing multilingual media monitoring
The rapid expansion of the number of different media sources has made news monitoring time consuming. Automatic screening of different newswires, radio and television channels and internet sites could therefore be a very useful service for many. Multilingual speech recognition technology can be combined with automatic structuring of audio and video data to lead to new developments for media watch applications. The EU-funded project has developed for demonstration a system that makes use of cutting-edge speech recognition technology to automatically detect specific topics in large databases of audio and video clips. After topic detection, it automatically indexes the broadcast data in four languages: English, French, German and Portuguese. The indexed data is then stored in a database for further use, such as e-mail notification of individuals interested in a specific topic or topics. Multilingual audio indexing thus makes media monitoring more efficient. Processing multilingual broadcast data for information access is of particular interest for media watch applications. In the case of major news events, similar topics are often covered by different media in different countries. The system created by the ALERT project picks up the same topic in different languages and brings the news to the individuals interested. The representation of the different data types in different languages uses a unified system. The audio and/or video segments matching a typed natural language query are selected from automatically derived transcriptions of programmes. These technologies could signal a new era in our ability to access news and staying up to date with topical developments.