On the Selection of the Best Retrieval Result Per Query - An Alternative Approach to Data Fusion