Forward-Backward

Summary

This is a dynamic programming algorithm, used in Hidden Markov Models to efficiently compute the posterior marginals over all the hidden state variables. This work extends IBM Model 1 and IBM Model 2, which models lexical translation probabilities and absolute distortion probabilities, by also modeling relative distortion.

The relative distortion is modeled by applying a first-order Hidden Markov Model, where each alignment probabilities are dependent on the distortion of the previous alignment.

Results indicate that Modeling the relative distortion can improve the overall quality of the Word Alignments.