Interested in Contributing?

Scored Systems

System Submitter System Notes Constraint Run Notes BLEU BLEU-cased TER BEER 2.0 CharactTER
Phrase-based MT System (Primary)  (Details) KIT
Karlsruhe Institute of Technology
Phrase-based SMT system Constraint condition using Giga Corpus no

30.5

29.5

0.582

LIGA-constrained-primary  (Details) LIGA
LIA/LIG collaboration
LIA/LIG Team French-English primary submission for contrained task no Primary Submission

29.9

28.8

0.584

LIGA  (Details) Marion
LIG
LIA/LIG collaboration no

29.9

28.7

0.585

RWTH-constrained-primary  (Details) mhuck
RWTH Aachen University
RWTH Aachen University French-English primary submission (constrained) no constrained, with gigaword

29.4

28.4

0.590

LIMSI-Ncode-Constrained-Primary  (Details) allauzen
LIMSI-CNRS
an open source statistical machine translation system based on bilingual n-grams. no Primary submission.

29.3

28.3

0.593

LIMSI-Ncode-Constrained-Constrast  (Details) allauzen
LIMSI-CNRS
an open source statistical machine translation system based on bilingual n-grams. no Constrastive experiments using paraphrases to increase the dev-set.

28.7

27.7

0.599

RWTH-Jane-constrained-contrastive  (Details) mhuck
RWTH Aachen University
RWTH Aachen University French-English Jane (hierarchical phrase-based) system, contrastive submission (constrained) no constrained, with gigaword

28.9

27.7

0.601

RWTH-PBT-constrained-contrastive  (Details) mhuck
RWTH Aachen University
RWTH Aachen University French-English PBT system, contrastive submission (constrained) no constrained, with gigaword

28.8

27.6

0.595

CMU Stat-XFER Group Primary  (Details) ghannema
Carnegie Mellon University
Joshua-based syntactic system also with non-syntactic phrases. Built from the WMT '11 data, including some Gigaword in the language model. Uses 102,000 selected grammar rules per test set. no Primary WMT '11 submission, constrained track.

27.8

26.9

0.593

CMU Stat-XFER Group Contrastive  (Details) ghannema
Carnegie Mellon University
Joshua-based syntactic system also with non-syntactic phrases. Built from the WMT '10 data (last year), with no Gigaword or Giga-FrEn. Uses 102,000 selected grammar rules per test set. no Contrastive WMT '11 submission, constrained track.

27.5

26.5

0.618

cmu-mm  (Details) mdenkows
Carnegie Mellon University
Constrained, publicly available data only. (Primary submission) no Optimized toward METEOR.

28.5

26.0

0.598

cmu-mb  (Details) mdenkows
Carnegie Mellon University
Constrained, publicly available data only. no Optimized toward BLEU.

28.4

25.8

0.600

Systran-RB+SPE  (Details) systran-user
Systran
Systran (rule-based) + statistical post edition, trained on monolingual (15M phrases) + 2009/10 news corpora no Systran (rule-based) + statistical post edition, trained on monolingual (15M phrases) + 2009/10 news corpora

26.7

25.4

0.614

jhu-hiero  (Details) jhu
Johns Hopkins University
Translation model: europarl and news, UN and 10^9 subsampled against dev (newstest2008) devtest (newstest2009) for MERT (700K sentences, 16M words) and test for decoding (1.2M sentences; about 20M words) in each language. Aligned using Giza++. Hiero grammar extracted with Thrax (http://github.com/jweese/thrax) with its standard 16 feature functions. 5-gram LM trained on the provided monolingual data except Gigaword (SRILM). MERT. Reranked 300-best using MBR. To recase, trained translation model with non-hierarchical rules of length up to 3, and a 5-gram cased LM. We submitted the 1-best recasing output detokenized with the provided script. no

28.0

25.2

0.598

uedin-wmt11-fr-en  (Details) Edinburgh
University of Edinburgh
no

25.8

24.8

0.631

uk-dan  (Details) zeman
Charles University in Prague, ┌FAL
no fr-en contrastive stc

24.5

22.1

0.651

seal.fr-en  (Details) sealPasaLab
NanJing University PASA lab
Nanjing University Wenjia Yang yangwen_jia@163.com, Lixin Liu, Min Chen, Rong Gu yes seal newstest2011 fr-en

26.1

20.2

0.625

uk-dan  (Details) zeman
Charles University in Prague, ┌FAL
no fr-en primary

24.8

19.3

0.647

LIUM FR-EN 2011  (Details) LIUM
LIUM, University of Le Mans
phrase-based system augmented with translations of the monolingual news corpus and a bitext retrieved by IR methods. Unknown words substitution as post-processing. no phrase-based system augmented with translations of the monolingual news corpus and a bitext retrieved by IR methods. Unknown words substitution as post-processing.

failed

failed

0.586