CimS – The CIS and IMS joint submission to WMT 2014 translating from English into German

Abstract

We present the CimS submissions to the 2014 Shared Task for the language pair EN→DE. We address the major problems that arise when translating into German: complex nominal and verbal morphology, productive compounding and flexible word ordering. Our morphologyaware translation systems handle word formation issues on different levels of morpho-syntactic modeling.

CimS - The CIS and IMS Joint Submission to WMT 2015 addressing morphological and syntactic differences in English to German SMT

“…However, the combination of nominal inflection and source-side reordering has a positive effect on translation quality. When it comes to the combination of compound processing and nominal inflection, which we have successfully applied in the past (Cap et al, 2014a;Cap et al, 2014b), we do not see any improvement in terms of BLEU score for this combination here. This does not necessarily mean that the compound systems quality is worse, as previous manual evaluations have shown that BLEU scores do not adequately reflect all compound-related improvements in translation quality (Cap et al, 2014a).…”

The LMU Munich Unsupervised Machine Translation Systems

“…Specifically, we participated in the unsupervised learning task which focuses on training MT models without access to any parallel data. The team has a strong track record at previous WMT shared tasks (Bojar et al, 2017(Bojar et al, , 2015(Bojar et al, , 2014(Bojar et al, , 2013 working on SMT systems (Cap et al, 2014(Cap et al, , 2015Weller et al, 2013;Peter et al, 2016; and proposed a top scoring linguistically informed neural machine translation system based on human evaluation at WMT17.…”

Large-scale Exact Decoding: The IMS-TTT submission to WMT14

“…We presented our submission to the WMT14 shared translation task based on a novel, promising "full syntax, no pruning" tree-to-tree approach to statistical machine translation, inspired by Huang 11 https://code.google.com/p/giza-pp/ 12 We use raw as described in (Cap et al, 2014b) as a fallback for RAW, RI for UNSPLIT and CoRI for SPLIT. et al (2006).…”

“…After translation, recasing was done by examining the output syntax tree, using a simple heuristics looking for nouns and sentence boundaries. Since coverage on the test set was also limited, we used the systems as described in (Cap et al, 2014b) 12 as a fallback to translate sentences that our system was not able to translate.…”