Interpolated backoff for factored translation models

Barry Haddow; Philipp Koehn

Interpolated backoff for factored translation models

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

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 language	English
Title of host publication	In Proceedings of the Tenth Conference of the Association for Machine Translation in the Americas (AMTA)
Publisher	Association for Machine Translation in the Americas, AMTA
Number of pages	10
Publication status	Published - 2012

Access to Document

http://amta2012.amtaweb.org/AMTA2012Files/papers/147.pdf

Cite this

@inproceedings{996f454bd2ac4bc89779ca158eedcdda,

title = "Interpolated backoff for factored translation models",

abstract = "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.",

author = "Barry Haddow and Philipp Koehn",

year = "2012",

language = "English",

booktitle = "In Proceedings of the Tenth Conference of the Association for Machine Translation in the Americas (AMTA)",

publisher = "Association for Machine Translation in the Americas, AMTA",

}

TY - GEN

T1 - Interpolated backoff for factored translation models

AU - Haddow, Barry

AU - Koehn, Philipp

PY - 2012

Y1 - 2012

N2 - 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.

AB - 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.

M3 - Conference contribution

BT - In Proceedings of the Tenth Conference of the Association for Machine Translation in the Americas (AMTA)

PB - Association for Machine Translation in the Americas, AMTA

ER -