A Bayesian framework for fusing multiple word knowledge models in videotext recognition | IEEE Conference Publication | IEEE Xplore