Zobrazeno 1 - 10
of 38
pro vyhledávání: '"Marcin Junczys-Dowmunt"'
We propose a novel scheme to use the Levenshtein Transformer to perform the task of word-level quality estimation. A Levenshtein Transformer is a natural fit for this task: trained to perform decoding in an iterative manner, a Levenshtein Transformer
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::edb07fdc30bd7eeff4c59e70a72e088b
http://arxiv.org/abs/2109.05611
http://arxiv.org/abs/2109.05611
Publikováno v:
NAACL-HLT
In this work, we study hallucinations in Neural Machine Translation (NMT), which lie at an extreme end on the spectrum of NMT pathologies. Firstly, we connect the phenomenon of hallucinations under source perturbation to the Long-Tail theory of Feldm
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::7e92d4793b150a6eb8776237733ff668
http://arxiv.org/abs/2104.06683
http://arxiv.org/abs/2104.06683
Autor:
Marcin Junczys-Dowmunt, Chris Hokamp, Fabio Natanael Kepler, André F. T. Martins, Roman Grundkiewicz, Ramón Fernandez Astudillo
Publikováno v:
Martins, A F T, Junczys-Dowmunt, M, Kepler, F N, Astudillo, R, Hokamp, C & Grundkiewicz, R 2017, ' Pushing the Limits of Translation Quality Estimation ', vol. 5, pp. 205-218 .
Translation quality estimation is a task of growing importance in NLP, due to its potential to reduce post-editing human effort in disruptive ways. However, this potential is currently limited by the relatively low accuracy of existing systems. In th
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::adc11cf6795a4d8b94c9f8822153000c
https://doi.org/10.1162/tacl_a_00056
https://doi.org/10.1162/tacl_a_00056
Publikováno v:
Grundkiewicz, R & Junczys-Dowmuntz, M 2019, Minimally-Augmented Grammatical Error Correction . in Proceedings of the 5th Workshop on Noisy User-generated Text (W-NUT 2019) . pp. 357–363, The 5th Workshop on Noisy User-generated Text (W-NUT): at EMNLP 2019, Hong Kong, 4/11/19 . https://doi.org/10.18653/v1/D19-5546
W-NUT@EMNLP
W-NUT@EMNLP
There has been an increased interest in low-resource approaches to automatic grammatical error correction. We introduce Minimally-Augmented Grammatical Error Correction (MAGEC) that does not require any error-labelled data. Our unsupervised approach
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::ec06b425bad07b999aca11080fce1058
https://hdl.handle.net/20.500.11820/eea4e69d-9333-484c-aebf-996eb16b362b
https://hdl.handle.net/20.500.11820/eea4e69d-9333-484c-aebf-996eb16b362b
Autor:
Alham Fikri Aji, Young Jin Kim, Roman Grundkiewicz, Kenneth Heafield, Marcin Junczys-Dowmunt, Hany Hassan, Nikolay Bogoychev
Publikováno v:
NGT@EMNLP-IJCNLP
Proceedings of the 3rd Workshop on Neural Generation and Translation
Kim, Y J, Junczys-Dowmunt, M, Hassan, H, Aji, A F, Heafield, K, Grundkiewicz, R & Bogoychev, N 2019, From Research to Production and Back: Ludicrously Fast Neural Machine Translation . in Proceedings of the The 3rd Workshop on Neural Generation and Translation (WNGT 2019) . Hong Kong, pp. 280–288, The 3rd Workshop on Neural Generation and Translation, Hong Kong, Hong Kong, 4/11/19 . https://doi.org/10.18653/v1/D19-5632
Proceedings of the 3rd Workshop on Neural Generation and Translation
Kim, Y J, Junczys-Dowmunt, M, Hassan, H, Aji, A F, Heafield, K, Grundkiewicz, R & Bogoychev, N 2019, From Research to Production and Back: Ludicrously Fast Neural Machine Translation . in Proceedings of the The 3rd Workshop on Neural Generation and Translation (WNGT 2019) . Hong Kong, pp. 280–288, The 3rd Workshop on Neural Generation and Translation, Hong Kong, Hong Kong, 4/11/19 . https://doi.org/10.18653/v1/D19-5632
This paper describes the submissions of the “Marian” team to the WNGT 2019 efficiency shared task. Taking our dominating submissions to the previous edition of the shared task as a starting point, we develop improved teacher-student training via
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::4b5768709a92eb40a46fed83d723003a
https://doi.org/10.18653/v1/d19-5632
https://doi.org/10.18653/v1/d19-5632
Publikováno v:
BEA@ACL
Grundkiewicz, R, Junczys-Dowmuntz, M & Heafield, K 2019, Neural Grammatical Error Correction Systems with Unsupervised Pre-training on Synthetic Data . in H Yannakoudakis, E Kochmar, C Leacock, N Madnani, I Pilán & T Zesch (eds), Proceedings of the Fourteenth Workshop on Innovative Use of NLP for Building Educational Applications ., W19-4427, Florence, Italy, pp. 252–263, 14th Workshop on Innovative Use of NLP for Building Educational Applications, Florence, Italy, 2/08/19 . < https://www.aclweb.org/anthology/W19-4427 >
University of Edinburgh-PURE
Grundkiewicz, R, Junczys-Dowmuntz, M & Heafield, K 2019, Neural Grammatical Error Correction Systems with Unsupervised Pre-training on Synthetic Data . in H Yannakoudakis, E Kochmar, C Leacock, N Madnani, I Pilán & T Zesch (eds), Proceedings of the Fourteenth Workshop on Innovative Use of NLP for Building Educational Applications ., W19-4427, Florence, Italy, pp. 252–263, 14th Workshop on Innovative Use of NLP for Building Educational Applications, Florence, Italy, 2/08/19 . < https://www.aclweb.org/anthology/W19-4427 >
University of Edinburgh-PURE
Considerable effort has been made to address the data sparsity problem in neural grammatical error correction. In this work, we propose a simple and surprisingly effective unsupervised synthetic error generation method based onconfusion sets extracte
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::cfe94fc34cc550db34e627b7c34e5d58
https://doi.org/10.18653/v1/w19-4427
https://doi.org/10.18653/v1/w19-4427
Autor:
Marcin Junczys-Dowmunt
Publikováno v:
WMT (2)
This paper describes the Microsoft Translator submissions to the WMT19 news translation shared task for English-German. Our main focus is document-level neural machine translation with deep transformer models. We start with strong sentence-level base
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::428db8c1f291fd2148eba07465c50a16
Publikováno v:
Bogoychev, N, Junczys-Dowmunt, M, Heafield, K & Aji, A 2018, Accelerating Asynchronous Stochastic Gradient Descent for Neural Machine Translation . in Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing . Brussels, Belgium, pp. 2991-2996, 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, 31/10/18 . < http://aclweb.org/anthology/D18-1332 >
University of Edinburgh-PURE
EMNLP
University of Edinburgh-PURE
EMNLP
In order to extract the best possible performance from asynchronous stochastic gradient descent one must increase the mini-batch size and scale the learning rate accordingly. In order to achieve further speedup we introduce a technique that delays gr
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::692104b82156a0d1cb132e7c613de28f
https://www.pure.ed.ac.uk/ws/files/75718984/Accelerating_Asynchronous_Stochastic_Gradient_Descent_for_Neural_Machine_Translation.pdf
https://www.pure.ed.ac.uk/ws/files/75718984/Accelerating_Asynchronous_Stochastic_Gradient_Descent_for_Neural_Machine_Translation.pdf
Publikováno v:
University of Edinburgh-PURE
Junczys-Dowmunt, M, Heafield, K, Hoang, H, Grundkiewicz, R & Aue, A 2018, Marian: Cost-effective High-Quality Neural Machine Translation in C++ . in Proceedings of the 2nd Workshop on Neural Machine Translation and Generation . pp. 129-135, 2nd Workshop on Neural Machine Translation and Generation, Melbourne, Victoria, Australia, 15/07/18 . < http://aclweb.org/anthology/W18-2716 >
NMT@ACL
Junczys-Dowmunt, M, Heafield, K, Hoang, H, Grundkiewicz, R & Aue, A 2018, Marian: Cost-effective High-Quality Neural Machine Translation in C++ . in Proceedings of the 2nd Workshop on Neural Machine Translation and Generation . pp. 129-135, 2nd Workshop on Neural Machine Translation and Generation, Melbourne, Victoria, Australia, 15/07/18 . < http://aclweb.org/anthology/W18-2716 >
NMT@ACL
This paper describes the submissions of the "Marian" team to the WNMT 2018 shared task. We investigate combinations of teacher-student training, low-precision matrix products, auto-tuning and other methods to optimize the Transformer model on GPU and
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::529790a7a8ea3ddf8bfb2535a2fbe1b5
http://arxiv.org/abs/1805.12096
http://arxiv.org/abs/1805.12096
Autor:
Ulrich Germann, Alham Fikri Aji, Marcin Junczys-Dowmunt, Tom Neckermann, Frank Seide, Kenneth Heafield, Tomasz Dwojak, Alexandra Birch, Roman Grundkiewicz, Nikolay Bogoychev, Hieu Hoang, André F. T. Martins
Publikováno v:
arXiv.org e-Print Archive
Junczys-Dowmunt, M, Grundkiewicz, R, Dwojak, T, Hoang, H, Heafield, K, Neckermann, T, Seide, F, Germann, U, Aji, A F, Bogoychev, N, Martins, A F T & Birch, A 2018, Marian: Fast Neural Machine Translation in C++ . in Proceedings of ACL 2018, System Demonstrations . pp. 116–121, 56th Annual Meeting of the Association for Computational Linguistics, Melbourne, Australia, 15/07/18 . https://doi.org/10.18653/v1/P18-4020
ACL (4)
Junczys-Dowmunt, M, Grundkiewicz, R, Dwojak, T, Hoang, H, Heafield, K, Neckermann, T, Seide, F, Germann, U, Aji, A F, Bogoychev, N, Martins, A F T & Birch, A 2018, Marian: Fast Neural Machine Translation in C++ . in Proceedings of ACL 2018, System Demonstrations . pp. 116–121, 56th Annual Meeting of the Association for Computational Linguistics, Melbourne, Australia, 15/07/18 . https://doi.org/10.18653/v1/P18-4020
ACL (4)
We present Marian, an efficient and self-contained Neural Machine Translation framework with an integrated automatic differentiation engine based on dynamic computation graphs. Marian is written entirely in C++. We describe the design of the encoder-
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::b8a4e0feaecdc54c56b8d2572462ec31
http://arxiv.org/abs/1804.00344
http://arxiv.org/abs/1804.00344