The increasing amount of information available in Internet makes human life more and more complicated. This huge amount of information especially in News pages makes people hard to recognize and follow new events happening in everyday life. In order to solve this problem, the news must be categorized according to events and their tracks automatically. New event detection and tracking, aims to solve this problem by automatically following news streams in order to detect the stories in news portals. We propose to adapt new event detection (NED) algorithms to Turkish test collection and see the effect of language difference in NED systems. Also, generally in news texts the important issues are discussed in the beginning of news; so with this in mind, we propose to apply a new technique that enhances similarity scoring techniques with term position information. This technique uses the position information of terms in a document to enhance weighting of terms. Lastly, we will use this technique to improve NED in English TDT-1 corpus.

DATE: 22 October, 2007, Monday@ 16:15
PLACE: EA 409