Advanced Temporal Reasoning for Precise Time-Stamping of Events (ARTEMIS)

Our work in Advanced Temporal Reasoning for Precise Time- Stamping of Events (ARTEMIS) addresses two challenges faced by systems which seek to enhance analysts’ awareness of event information. First, while today’s text processing systems have greatly increased the amount of temporal information available from a document collection, most systems have focused on extracting only the temporal information that is overtly mentioned in a text. We demonstrate that our approach can acquire all of the essential temporal information associated with events. In addition to a state-of-the-art TERN system, we use a combination of approaches to acquire the knowledge needed to infer implicit forms of temporal information. Second, in order to take advantage of the explosion of temporal information that we will make available, event we will provide temporal reasoning capabilities which will allow event visualization systems to “make sense” of complex situations automatically. The cornerstone of our Phase I work will be a graph-based temporal reasoning component which can track changes in reference time throughout a document, estimate the duration of specific events, identify the likely order of a sequence of events, and compute an exact time-stamp for events which are not associated with overt time or date information.