An enriched MT grammar for under $100

Authors:
Omar F. Zaidan;Juri Ganitkevitch
Affiliations:
Johns Hopkins University, Baltimore, MD;Johns Hopkins University, Baltimore, MD
Venue:
CSLDAMT '10 Proceedings of the NAACL HLT 2010 Workshop on Creating Speech and Language Data with Amazon's Mechanical Turk
Year:
2010

Citing 3
Cited 1

Moses: open source toolkit for statistical machine translation

ACL '07 Proceedings of the 45th Annual Meeting of the ACL on Interactive Poster and Demonstration Sessions
Joshua: an open source toolkit for parsing-based machine translation

StatMT '09 Proceedings of the Fourth Workshop on Statistical Machine Translation
Syntax augmented machine translation via chart parsing

StatMT '06 Proceedings of the Workshop on Statistical Machine Translation

Creating speech and language data with Amazon's Mechanical Turk

CSLDAMT '10 Proceedings of the NAACL HLT 2010 Workshop on Creating Speech and Language Data with Amazon's Mechanical Turk

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose a framework for improving output quality of machine translation systems, by operating on the level of grammar rule features. Our framework aims to give a boost to grammar rules that appear in the derivations of translation candidates that are deemed to be of good quality, hence making those rules more preferable by the system. To that end, we ask human annotators on Amazon Mechanical Turk to compare translation candidates, and then interpret their preferences of one candidate over another as an implicit preference for one derivation over another, and therefore as an implicit preference for one or more grammar rules. Our framework also allows us to generalize these preferences to grammar rules corresponding to a previously unseen test set, namely rules for which no candidates have been judged.