Friday, June 25, 1pm, 2405 Siebel Center (NOTE ROOM CHANGE)

“Relationship Extraction for Event Summarization”

Dr. William Hsu, Kansas State University

This talk will survey current open problems and approaches used in the domain of relationship extraction for summarization of events.  I will first present several existing techniques for spatiotemporal event extraction and discuss the challenge of applying these techniques to domains such as epidemiology and health informatics.  Information extraction tasks in this area are related to update summarization, the problem of automatically generating a brief restatement of the main points in a text when the user of the system has already read a given set of earlier articles.  Next, I will survey some of the machine learning and natural language processing methodologies used in this type of event summarization.  I will then describe the related task of topic detection and tracking with application to tracking current events, and discuss some relevant current work on topic modeling for this task.

Finally, I will present some results from continuing research on development of analytical tools based on this work, the emphasis of which is on performance elements such as predictive data mining, data-aware search, and question answering.  I will conclude with a discussion of relevant related work in thematic mapping of relationships and opinions, focusing on link analysis approaches.