Towards Automatic Event Extraction

Abstract. An automatic news tracking and analysis system which records world events over long time periods is described. It allows to track country spe-cific news, the activities of individual persons and groups, to derive trends, and to provide data for further analysis and research. The data source is the Europe Media Monitor (EMM) which monitors news from around the world in real time via the Internet and from various News Agencies. EMM’s main pur-pose is to provide rapid feedback of press coverage and breaking news for European Policy Makers. Increasingly, however it is being used for security ap-plications and for foreign policy monitoring. This paper describes how lan-guage technologies and clustering techniques have been applied to the 30,000 daily news reports to derive the top stories in each of 13 languages, to locate events geospatially, and to extract and record entities involved. Related stories have been linked across time and across languages, allowing for national com-parisons and to derive name variants. Results and future plans are described.JRC.G.2-Support to external securit