This small dataset contains 3 speech corpora collected using the Alex Translate telephone service (https://ufal.mff.cuni.cz/alex#alex-translate).
The "part1" and "part2" corpora contain English speech with transcriptions and Czech translations. These recordings were collected from users of the service. Part 1 contains earlier recordings, filtered to include only clean speech; Part 2 contains later recordings with no filtering applied.
The "cstest" corpus contains recordings of artificially created sentences, each containing one or more Czech names of places in the Czech Republic. These were recorded by a multinational group of students studying in Prague.
Human post-edited test sentences for the WMT 2017 Automatic post-editing task. This consists in 2,000 English sentences belonging to the IT domain and already tokenized. Source and target segments can be downloaded from: https://lindat.mff.cuni.cz/repository/xmlui/handle/11372/LRT-2132. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Human post-edited test sentences for the WMT 2017 Automatic post-editing task. This consists in 2,000 German sentences belonging to the IT domain and already tokenized. Source and target segments can be downloaded from: https://lindat.mff.cuni.cz/repository/xmlui/handle/11372/LRT-2133. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Human post-edited and reference test sentences for the En-De PBSMT WMT 2018 Automatic post-editing task. This consists of 2,000 German sentences for each file belonging to the IT domain and already tokenized. All data is provided by the EU project QT21 (http://www.qt21.eu/).
CsEnVi Pairwise Parallel Corpora consist of Vietnamese-Czech parallel corpus and Vietnamese-English parallel corpus. The corpora were assembled from the following sources:
- OPUS, the open parallel corpus is a growing multilingual corpus of translated open source documents.
The majority of Vi-En and Vi-Cs bitexts are subtitles from movies and television series.
The nature of the bitexts are paraphrasing of each other's meaning, rather than translations.
- TED talks, a collection of short talks on various topics, given primarily in English, transcribed and with transcripts translated to other languages. In our corpus, we use 1198 talks which had English and Vietnamese transcripts available and 784 talks which had Czech and Vietnamese transcripts available in January 2015.
The size of the original corpora collected from OPUS and TED talks is as follows:
CS/VI EN/VI
Sentence 1337199/1337199 2035624/2035624
Word 9128897/12073975 16638364/17565580
Unique word 224416/68237 91905/78333
We improve the quality of the corpora in two steps: normalizing and filtering.
In the normalizing step, the corpora are cleaned based on the general format of subtitles and transcripts. For instance, sequences of dots indicate explicit continuation of subtitles across multiple time frames. The sequences of dots are distributed differently in the source and the target side. Removing the sequence of dots, along with a number of other normalization rules, improves the quality of the alignment significantly.
In the filtering step, we adapt the CzEng filtering tool [1] to filter out bad sentence pairs.
The size of cleaned corpora as published is as follows:
CS/VI EN/VI
Sentence 1091058/1091058 1113177/1091058
Word 6718184/7646701 8518711/8140876
Unique word 195446/59737 69513/58286
The corpora are used as training data in [2].
References:
[1] Ondřej Bojar, Zdeněk Žabokrtský, et al. 2012. The Joy of Parallelism with CzEng 1.0. Proceedings of LREC2012. ELRA. Istanbul, Turkey.
[2] Duc Tam Hoang and Ondřej Bojar, The Prague Bulletin of Mathematical Linguistics. Volume 104, Issue 1, Pages 75–86, ISSN 1804-0462. 9/2015
CUBBITT En-Cs translation models, exported via TensorFlow Serving, available in the Lindat translation service (https://lindat.mff.cuni.cz/services/translation/).
Models are compatible with Tensor2tensor version 1.6.6.
For details about the model training (data, model hyper-parameters), please contact the archive maintainer.
Evaluation on newstest2014 (BLEU):
en->cs: 27.6
cs->en: 34.4
(Evaluated using multeval: https://github.com/jhclark/multeval)
CUBBITT En-Fr translation models, exported via TensorFlow Serving, available in the Lindat translation service (https://lindat.mff.cuni.cz/services/translation/).
Models are compatible with Tensor2tensor version 1.6.6.
For details about the model training (data, model hyper-parameters), please contact the archive maintainer.
Evaluation on newstest2014 (BLEU):
en->fr: 38.2
fr->en: 36.7
(Evaluated using multeval: https://github.com/jhclark/multeval)
CUBBITT En-Pl translation models, exported via TensorFlow Serving, available in the Lindat translation service (https://lindat.mff.cuni.cz/services/translation/).
Models are compatible with Tensor2tensor version 1.6.6.
For details about the model training (data, model hyper-parameters), please contact the archive maintainer.
Evaluation on newstest2020 (BLEU):
en->pl: 12.3
pl->en: 20.0
(Evaluated using multeval: https://github.com/jhclark/multeval)
This package contains the eye-tracker recordings of 8 subjects evaluating English-to-Czech machine translation quality using the WMT-style ranking of sentences.
We provide the set of sentences evaluated, the exact screens presented to the annotators (including bounding box information for every area of interest and even for individual letters in the text) and finally the raw EyeLink II files with gaze trajectories.
The description of the experiment can be found in the paper:
Ondřej Bojar, Filip Děchtěrenko, Maria Zelenina. A Pilot Eye-Tracking Study of WMT-Style Ranking Evaluation.
Proceedings of the LREC 2016 Workshop “Translation Evaluation – From Fragmented Tools
and Data Sets to an Integrated Ecosystem”, Georg Rehm, Aljoscha Burchardt et al. (eds.). pp. 20-26. May 2016, Portorož, Slovenia.
This work has received funding from the European Union's Horizon 2020 research
and innovation programme under grant agreement no. 645452 (QT21). This work was
partially financially supported by the Government of Russian Federation, Grant
074-U01.
This work has been using language resources developed, stored and distributed
by the LINDAT/CLARIN project of the Ministry of Education, Youth and Sports of
the Czech Republic (project LM2010013).
Post-editing and MQM annotations produced by the QT21 project. As described in
@InProceedings{specia-etal_MTSummit:2017,
author = {Specia, Lucia and Kim Harris and Frédéric Blain and Aljoscha Burchardt and Viviven Macketanz and Inguna Skadiņa and Matteo Negri and and Marco Turchi},
title = {Translation Quality and Productivity: A Study on Rich Morphology Languages},
booktitle = {Proceedings of Machine Translation Summit XVI},
year = {2017},
pages = {55--71},
address = {Nagoya, Japan},
}
Test data for the WMT 2017 Automatic post-editing task (the same used for the Sentence-level Quality Estimation task). They consist in German-English triplets (source and target) belonging to the pharmacological domain and already tokenized. Test set contains 2,000 pairs. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Test data for the WMT 2017 Automatic post-editing task (the same used for the Sentence-level Quality Estimation task). They consist in 2,000 English-German pairs (source and target) belonging to the IT domain and already tokenized. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Test data for the WMT 2018 Automatic post-editing task. They consist in English-German pairs (source and target) belonging to the information technology domain and already tokenized. Test set contains 1,023 pairs. A neural machine translation system has been used to generate the target segments. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Test data for the WMT 2018 Automatic post-editing task. They consist in English-German pairs (source and target) belonging to the information technology domain and already tokenized. Test set contains 2,000 pairs. A phrase-based machine translation system has been used to generate the target segments. This test set is sampled from the same dataset used for the 2016 and 2017 APE shared task editions. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Training, development and text data (the same used for the Sentence-level Quality Estimation task) consist in English-German triplets (source, target and post-edit) belonging to the IT domain and already tokenized.
Training and development respectively contain 12,000 and 1,000 triplets, while the test set 2,000 instances. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Training, development and test data consist in German sentences belonging to the IT domain and already tokenized. These sentences are the references of the data released for the 2016 edition of the WMT APE shared task. Differently from the data previously released, these sentences are obtained by manually translating the source sentence without leveraging the raw mt outputs. Training and development respectively contain 12,000 and 1,000 segments, while the test set 2,000 items. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Training and development data for the WMT16 QE task. Test data will be published as a separate item.
This shared task will build on its previous four editions to further examine automatic methods for estimating the quality of machine translation output at run-time, without relying on reference translations. We include word-level, sentence-level and document-level estimation. The sentence and word-level tasks will explore a large dataset produced from post-editions by professional translators (as opposed to crowdsourced translations as in the previous year). For the first time, the data will be domain-specific (IT domain). The document-level task will use, for the first time, entire documents, which have been human annotated for quality indirectly in two ways: through reading comprehension tests and through a two-stage post-editing exercise. Our tasks have the following goals:
- To advance work on sentence and word-level quality estimation by providing domain-specific, larger and professionally annotated datasets.
- To study the utility of detailed information logged during post-editing (time, keystrokes, actual edits) for different levels of prediction.
- To analyse the effectiveness of different types of quality labels provided by humans for longer texts in document-level prediction.
This year's shared task provides new training and test datasets for all tasks, and allows participants to explore any additional data and resources deemed relevant. A in-house MT system was used to produce translations for the sentence and word-level tasks, and multiple MT systems were used to produce translations for the document-level task. Therefore, MT system-dependent information will be made available where possible.
The item contains models to tune for the WMT16 Tuning shared task for Czech-to-English.
CzEng 1.6pre (http://ufal.mff.cuni.cz/czeng/czeng16pre) corpus is used for the training of the translation models. The data is tokenized (using Moses tokenizer), lowercased and sentences longer than 60 words and shorter than 4 words are removed before training. Alignment is done using fast_align (https://github.com/clab/fast_align) and the standard Moses pipeline is used for training.
Two 5-gram language models are trained using KenLM: one only using the CzEng English data and the other is trained using all available English mono data for WMT except Common Crawl.
Also included are two lexicalized bidirectional reordering models, word based and hierarchical, with msd conditioned on both source and target of processed CzEng.
This item contains models to tune for the WMT16 Tuning shared task for English-to-Czech.
CzEng 1.6pre (http://ufal.mff.cuni.cz/czeng/czeng16pre) corpus is used for the training of the translation models. The data is tokenized (using Moses tokenizer), lowercased and sentences longer than 60 words and shorter than 4 words are removed before training. Alignment is done using fast_align (https://github.com/clab/fast_align) and the standard Moses pipeline is used for training.
Two 5-gram language models are trained using KenLM: one only using the CzEng Czech data and the other is trained using all available Czech mono data for WMT except Common Crawl.
Also included are two lexicalized bidirectional reordering models, word based and hierarchical, with msd conditioned on both source and target of processed CzEng.
Training and development data for the WMT 2017 Automatic post-editing task (the same used for the Sentence-level Quality Estimation task). They consist in German-English triplets (source, target and post-edit) belonging to the pharmacological domain and already tokenized. Training and development respectively contain 25,000 and 1,000 triplets. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Training data for the WMT 2017 Automatic post-editing task (the same used for the Sentence-level Quality Estimation task). They consist in 11,000 English-German triplets (source, target and post-edit) belonging to the IT domain and already tokenized. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Training and development data for the WMT17 QE task. Test data will be published as a separate item.
This shared task will build on its previous five editions to further examine automatic methods for estimating the quality of machine translation output at run-time, without relying on reference translations. We include word-level, phrase-level and sentence-level estimation. All tasks will make use of a large dataset produced from post-editions by professional translators. The data will be domain-specific (IT and Pharmaceutical domains) and substantially larger than in previous years. In addition to advancing the state of the art at all prediction levels, our goals include:
- To test the effectiveness of larger (domain-specific and professionally annotated) datasets. We will do so by increasing the size of one of last year's training sets.
- To study the effect of language direction and domain. We will do so by providing two datasets created in similar ways, but for different domains and language directions.
- To investigate the utility of detailed information logged during post-editing. We will do so by providing post-editing time, keystrokes, and actual edits.
This year's shared task provides new training and test datasets for all tasks, and allows participants to explore any additional data and resources deemed relevant. A in-house MT system was used to produce translations for all tasks. MT system-dependent information can be made available under request. The data is publicly available but since it has been provided by our industry partners it is subject to specific terms and conditions. However, these have no practical implications on the use of this data for research purposes.
Test data for the WMT17 QE task. Train data can be downloaded from http://hdl.handle.net/11372/LRT-1974
This shared task will build on its previous five editions to further examine automatic methods for estimating the quality of machine translation output at run-time, without relying on reference translations. We include word-level, phrase-level and sentence-level estimation. All tasks will make use of a large dataset produced from post-editions by professional translators. The data will be domain-specific (IT and Pharmaceutical domains) and substantially larger than in previous years. In addition to advancing the state of the art at all prediction levels, our goals include:
- To test the effectiveness of larger (domain-specific and professionally annotated) datasets. We will do so by increasing the size of one of last year's training sets.
- To study the effect of language direction and domain. We will do so by providing two datasets created in similar ways, but for different domains and language directions.
- To investigate the utility of detailed information logged during post-editing. We will do so by providing post-editing time, keystrokes, and actual edits.
This year's shared task provides new training and test datasets for all tasks, and allows participants to explore any additional data and resources deemed relevant. A in-house MT system was used to produce translations for all tasks. MT system-dependent information can be made available under request. The data is publicly available but since it has been provided by our industry partners it is subject to specific terms and conditions. However, these have no practical implications on the use of this data for research purposes.
Training and development data for the WMT 2018 Automatic post-editing task. They consist in English-German triplets (source, target and post-edit) belonging to the information technology domain and already tokenized. Training and development respectively contain 13,442 and 1,000 triplets. A neural machine translation system has been used to generate the target segments. All data is provided by the EU project QT21 (http://www.qt21.eu/).
Test data for the WMT18 QE task. Train data can be downloaded from http://hdl.handle.net/11372/LRT-2619.
This shared task will build on its previous six editions to further examine automatic methods for estimating the quality of machine translation output at run-time, without relying on reference translations. We include word-level, phrase-level and sentence-level estimation. All tasks make use of datasets produced from post-editions by professional translators. The datasets are domain-specific (IT and life sciences/pharma domains) and extend from those used previous years with more instances and more languages. One important addition is that this year we also include datasets with neural MT outputs. In addition to advancing the state of the art at all prediction levels, our specific goals are:
To study the performance of quality estimation approaches on the output of neural MT systems. We will do so by providing datasets for two language language pairs where the same source segments are translated by both a statistical phrase-based and a neural MT system.
To study the predictability of deleted words, i.e. words that are missing in the MT output. TO do so, for the first time we provide data annotated for such errors at training time.
To study the effectiveness of explicitly assigned labels for phrases. We will do so by providing a dataset where each phrase in the output of a phrase-based statistical MT system was annotated by human translators.
To study the effect of different language pairs. We will do so by providing datasets created in similar ways for four language language pairs.
To investigate the utility of detailed information logged during post-editing. We will do so by providing post-editing time, keystrokes, and actual edits.
Measure progress over years at all prediction levels. We will do so by using last year's test set for comparative experiments.
In-house statistical and neural MT systems were built to produce translations for all tasks. MT system-dependent information can be made available under request. The data is publicly available but since it has been provided by our industry partners it is subject to specific terms and conditions. However, these have no practical implications on the use of this data for research purposes. Participants are allowed to explore any additional data and resources deemed relevant.
Training and development data for the WMT18 QE task. Test data will be published as a separate item.
This shared task will build on its previous six editions to further examine automatic methods for estimating the quality of machine translation output at run-time, without relying on reference translations. We include word-level, phrase-level and sentence-level estimation. All tasks make use of datasets produced from post-editions by professional translators. The datasets are domain-specific (IT and life sciences/pharma domains) and extend from those used previous years with more instances and more languages. One important addition is that this year we also include datasets with neural MT outputs. In addition to advancing the state of the art at all prediction levels, our specific goals are:
To study the performance of quality estimation approaches on the output of neural MT systems. We will do so by providing datasets for two language language pairs where the same source segments are translated by both a statistical phrase-based and a neural MT system.
To study the predictability of deleted words, i.e. words that are missing in the MT output. TO do so, for the first time we provide data annotated for such errors at training time.
To study the effectiveness of explicitly assigned labels for phrases. We will do so by providing a dataset where each phrase in the output of a phrase-based statistical MT system was annotated by human translators.
To study the effect of different language pairs. We will do so by providing datasets created in similar ways for four language language pairs.
To investigate the utility of detailed information logged during post-editing. We will do so by providing post-editing time, keystrokes, and actual edits.
Measure progress over years at all prediction levels. We will do so by using last year's test set for comparative experiments.
In-house statistical and neural MT systems were built to produce translations for all tasks. MT system-dependent information can be made available under request. The data is publicly available but since it has been provided by our industry partners it is subject to specific terms and conditions. However, these have no practical implications on the use of this data for research purposes. Participants are allowed to explore any additional data and resources deemed relevant.