The sum of information on the internet has been growing fast and this threatens to overwhelm human attention and raises a new challenge for statistical text classification in tracking of specified events from chronologically ordered document streams. The tracking task was defined by the Topic Detection and Tracking (TDT) research initiative and the task is a supervised process that is similar in nature to the problem information filtering, routing and filtering and categorization. The number of positive instances per events is extremely is small, the majority of training documents are unlabelled, and most of the events have a short duration in time and this makes event tracking extremely difficult. We proposed several supervised text categorization methods, specifically new variants of Rocchio approach, and new variants of k-nearest Neighbor (kNN) algorithm to track the events. These algorithms are performed on Turkish Test Collection to see their effectiveness.

DATE: 22 October, 2007, Monday@ 16:50
PLACE: EA 409