Automatically Time Stamping Events in Unrestricted Text

This proposal addresses the design and development of a temporal model to be incorporated as part of an information extraction system. The model will enable time stamping of critical information such as relationships and events; this in turns enablessophisticated visualization of information on timelines. Advanced temporal reasoning, including tracking of reference time shifts will be pursued. The Phase 1 effort will focus on the design of the temporal model, as well as its use in some sample timestamping tasks. The Phase 2 effort will focus on enhanced capability of the temporal model, as well as integration into an existing information extractionengine.Exact information about the date and time of an event is often not available from the text. However, it is usually possible for human readers to construct a partial order of events, and this is sufficient to understand the text. A key goal of this projectis to support the processing of event information from texts even when only ordering information is available from the sources. Information analysts and decision makers will benefit since the proposed effort extends the utility of current information extraction technology. Without time stamping of events, or at least some temporal ordering, event information cannot be used inapplications such as visualization. A commercially viable solution has many applications. The ability to gather intelligence from large volumes of publicly available information (newspapers, websites, newsgroups, etc.) is invaluable. Automaticallyderiving useful information is based on detection of key events, which in turn requires time stamping. Furthermore, business intelligence systems use large knowledge-bases of companies, products, people and projects. Updating these knowledge-bases fromdiverse text sources in an automatic or semi-automatic manner would be valuable; time stamped event detection is a key to this. Our commercial application focused on brand perception and monitoring will benefit. Military applications stem from the factthat analysts need assistance in visualizing events in order to detect hidden trends and patterns in large volumes of text.