DATASETS.md

June 19, 2024 · View on GitHub

DatasetDescription
mtedx/validmTEDx evaluation data, valid: URL
mtedx/testmTEDx evaluation data, test: URL
wmt23Official evaluation and system data for WMT23.
wmt22Official evaluation and system data for WMT22.
wmt21/systemsWMT21 system output.
wmt21/devDevelopment data for WMT21，if multiple references are available, the first one is used.
wmt21/DOfficial evaluation data for WMT21 with reference D
wmt21/COfficial evaluation data for WMT21 with reference C
wmt21/BOfficial evaluation data for WMT21 with reference B.
wmt21/ACOfficial evaluation data for WMT21 with references A and C
wmt21/ABOfficial evaluation data for WMT21 with references A and B.
wmt21Official evaluation data for WMT21.
wmt20/robust/set1WMT20 robustness task, set 1
wmt20/robust/set2WMT20 robustness task, set 2
wmt20/robust/set3WMT20 robustness task, set 3
wmt20/tworefsWMT20 news test sets with two references
wmt20Official evaluation data for WMT20
mtnt2019Test set for the WMT 19 robustness shared task
mtnt1.1/testTest data for the Machine Translation of Noisy Text task: URL
mtnt1.1/validValidation data for the Machine Translation of Noisy Text task: URL
mtnt1.1/trainTraining data for the Machine Translation of Noisy Text task: URL
wmt20/devDevelopment data for tasks new to 2020.
wmt19Official evaluation data.
wmt19/devDevelopment data for tasks new to 2019.
wmt19/google/arAdditional high-quality reference for WMT19/en-de.
wmt19/google/arpAdditional paraphrase of wmt19/google/ar.
wmt19/google/wmtpAdditional paraphrase of the official WMT19 reference.
wmt19/google/hqrBest human selected-reference between wmt19 and wmt19/google/ar.
wmt19/google/hqpBest human-selected reference between wmt19/google/arp and wmt19/google/wmtp.
wmt19/google/hqallBest human-selected reference among original official reference and the Google reference and paraphrases.
wmt18Official evaluation data.
wmt18/test-tsOfficial evaluation sources with extra test sets interleaved.
wmt18/devDevelopment data (Estonian<>English).
wmt17Official evaluation data.
wmt17/BAdditional reference for EN-FI and FI-EN.
wmt17/tworefsSystems with two references.
wmt17/improvedImproved zh-en and en-zh translations.
wmt17/devDevelopment sets released for new languages in 2017.
wmt17/msAdditional Chinese-English references from Microsoft Research.
wmt16Official evaluation data.
wmt16/BAdditional reference for EN-FI.
wmt16/tworefsEN-FI with two references.
wmt16/devDevelopment sets released for new languages in 2016.
wmt15Official evaluation data.
wmt14Official evaluation data.
wmt14/fullEvaluation data released after official evaluation for further research.
wmt13Official evaluation data.
wmt12Official evaluation data.
wmt11Official evaluation data.
wmt10Official evaluation data.
wmt09Official evaluation data.
wmt08Official evaluation data.
wmt08/ncOfficial evaluation data (news commentary).
wmt08/europarlOfficial evaluation data (Europarl).
iwslt17Official evaluation data for IWSLT.
iwslt17/tst2016Development data for IWSLT 2017.
iwslt17/tst2015Development data for IWSLT 2017.
iwslt17/tst2014Development data for IWSLT 2017.
iwslt17/tst2013Development data for IWSLT 2017.
iwslt17/tst2012Development data for IWSLT 2017.
iwslt17/tst2011Development data for IWSLT 2017.
iwslt17/tst2010Development data for IWSLT 2017.
iwslt17/dev2010Development data for IWSLT 2017.
multi30k/20162016 flickr test set of Multi30k dataset
multi30k/20172017 flickr test set of Multi30k dataset
multi30k/20182018 flickr test set of Multi30k dataset. See URL for evaluation.