Interpolated backoff for factored translation models

Barry Haddow, Philipp Koehn

Research output: Chapter in Book/Report/Conference proceedingConference contribution


We propose interpolated backoff methods to strike the balance between traditional surface form translation models and factored models that decompose translation into lemma and morphological feature mapping steps. We show that this approach improves translation quality by 0.5 BLEU (German–English) over phrase-based models, due to the better translation of rare nouns and adjectives.
Original languageEnglish
Title of host publicationIn Proceedings of the Tenth Conference of the Association for Machine Translation in the Americas (AMTA)
PublisherAssociation for Machine Translation in the Americas, AMTA
Number of pages10
Publication statusPublished - 2012

Cite this