Phrasetable smoothing for statistical machine translation

Abstract

We discuss different strategies for smoothing the phrasetable in Statistical MT, and give results over a range of translation settings. We show that any type of smoothing is a better idea than the relativefrequency estimates that are often used. The best smoothing techniques yield consistent gains of approximately 1% (absolute) according to the BLEU metric.

Edinburgh's Statistical Machine Translation Systems for WMT16

“…log probability), phrase translation and lexical translation scores, word and phrase penalties, and a linear distortion score. The phrase translation probabilities are smoothed with Good-Turing smoothing (Foster et al, 2006). We used the hierarchical lexicalized reordering model (Galley and Manning, 2008) with 4 possible orientations (monotone, swap, discontinuous left and discontinuous right) in both left-to-right and right-to-left direction.…”

The Edinburgh/JHU Phrase-based Machine Translation Systems for WMT~2015

“…The core features of our model are a 5-gram LM score, phrase translation and lexical translation scores, word and phrase penalties, and a linear distortion score. The phrase translation probabilities are smoothed with GoodTuring smoothing (Foster et al, 2006). We used the hierarchical lexicalized reordering model (Galley and Manning, 2008) with 4 possible orientations (monotone, swap, discontinuous left and discontinuous right) in both left-to-right and rightto-left direction.…”

GREAT: open source software for statistical machine translation

“…Smoothing is still an open problem in SMT (Foster et al, 2006) and SFSTs are not an exception: they are not completely smoothed just by smoothing the corresponding SFSAs. Future research on SFST smoothing should be based on specific criteria for SFSTs as well.…”