Incorporating Position Information into a Maximum Entropy/Minimum Divergence Translation Model