file.pdf (325.25 kB)
Download file

A Simple, Fast, and Effective Reparameterization of IBM Model 2

Download (325.25 kB)
journal contribution
posted on 01.06.2013, 00:00 authored by Chris Dyer, Victor Chahuneau, Noah A. Smith

We present a simple log-linear reparameterization of IBM Model 2 that overcomes problems arising from Model 1’s strong assumptions and Model 2’s overparameterization. Efficient inference, likelihood evaluation, and parameter estimation algorithms are provided. Training the model is consistently ten times faster than Model 4. On three large-scale translation tasks, systems built using our alignment model outperform IBM Model 4.

An open-source implementation of the alignment model described in this paper is available from align .


Publisher Statement

Copyright 2013 ACL



Usage metrics