... yi. The features in the model are up-dated, and the algorithm moves to the next utterance.After each pass over the training data, performance on a held-out data set is evaluated, and the parameterizationwith ... discriminative language modelingfor a large vocabulary speech recognition task. We con-trast two parameter estimation methods: the perceptronalgorithm, and a method based on conditional randomfields ... what effect each parame-ter has on the error rate, and then modifies the parametersto reduce the error rate based on this prediction.2 Linear Models, the PerceptronAlgorithm, and Conditional...