ABSTRACT
Weighted indexing efficiently used in text retrieval systems has recently been proposed for content-based indexing of sports news videos. A sports news video is then indexed by a set of the weighted sports categories reported in this news. The temporal aggregation reduces analyses to player scenes the most suitable for scene categorization. The high similarities measured by the Euclidian and Chebyshew distances of weighting results obtained for nonaggregated and aggregated videos of the digital video test collection confirm that weighted indexing can be also applied to temporally aggregated sports news videos without fear that the aggregation distorts final weighting results.