Abstract

We’re going to talk about statistical machine translation (SMT). We’ll cover some basic SMT models, outline the improvements that have turned out to give the biggest gains, and talk about two extensions in more detail.

Firstly we’ll discuss how MapReduce can be used for SMT :

T. Brants, A. Popat, P. Xu, F. Och, and J. Dean, “Large Language Models in Machine Translation,” Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (EMNLP-CoNLL), 2007, pp. 867, 858.

C. Dyer, A. Cordova, A. Mont, and J. Lin, “Fast, easy, and cheap: Construction of statistical machine translation models with MapReduce,” Proceedings of the Third Workshop on Statistical Machine Translation, 2008, pp. 199–207.

Then we’ll talk about hierarchical phrase-based translation, which is theoretically pretty as well as giving significant improvements in translation:

Chiang (2005) – “A hierarchical phrase-based model for statistical machine translation” (http://acl.ldc.upenn.edu/P/P05/p05-1033.pdf)

G. Iglesias, A. de Gispert, E. R. Banga and W. Byrne. (2009) “Hierarchical Phrase-Based Translation with Weighted Finite State Transducers” (http://www.aclweb.org/anthology-new/N/N09/N09-1049.pdf)

Background material we’ve found helpful includes the book “Statistical Machine Translation” by Koehn (2009) (http://www.statmt.org/), and the review paper by Lopez (2008) “Statistical Machine Translation” (http://homepages.inf.ed.ac.uk/alopez/papers/survey.pdf).