Alignment by Agreement

Abstract

We present an unsupervised approach to symmetric word alignment in which two simple asymmetric models are trained jointly to maximize a combination of data likelihood and agreement between the models. Compared to the standard practice of intersecting predictions of independently-trained models, joint training provides a 32% reduction in AER. Moreover, a simple and efficient pair of HMM aligners provides a 29% reduction in AER over symmetrized IBM model 4 predictions.

Extracted Key Phrases

5 Figures and Tables

0204060'05'06'07'08'09'10'11'12'13'14'15'16'17
Citations per Year

445 Citations

Semantic Scholar estimates that this publication has 445 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@inproceedings{Liang2006AlignmentBA, title={Alignment by Agreement}, author={Percy Liang and Ben Taskar and Dan Klein}, booktitle={HLT-NAACL}, year={2006} }