Pre-trained language model representations for language generation
Open accessAbstract
Sergey Edunov, Alexei Baevski, Michael Auli. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Previous work has shown that models can be improved by incorporating pre-trained encoders, but gains from using pre-trained language models in decoders have been limited.
“Previous work Edunov et al. (2019) has shown that models can be improved by incorporating pre-trained encoders, but gains from using pre-trained language models in decoders have been limited.”
From BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension (2020) · cites this paper as Edunov et al. (2019)
Other work has shown that encoders can be improved using pre-trained representations, but gains in decoders are more limited.
“Other work has shown that encoders can be improved using pre-trained representations (Edunov et al., 2019), but gains in decoders are more limited.”
From BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension (2020) · cites this paper as Edunov et al., 2019
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · see citing works
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 48,293 · see citing works
BLEU: a method for automatic evaluation of machine translation
Papineni, K., Roukos, S., Ward, T., & Zhu, W.-J. (2001) · Proceedings of the 40th Annual Meeting on Association for Computational Linguistics - ACL '02 · doi:10.3115/1073083.1073135 · Cited by 21,344 · see citing works
Deep Contextualized Word Representations
Peters, M., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1202 · Cited by 11,983 · see citing works
Moses: open source toolkit for statistical machine translation
Koehn, P., Zens, R., Dyer, C., Bojar, O., Constantin, A., Herbst, E., Hoang, H., Birch, A., Callison-Burch, C., Federico, M., et al. (2007) · Proceedings of the 45th Annual Meeting of the ACL on Interactive Poster and Demonstration Sessions - ACL '07 · doi:10.3115/1557769.1557821 · Cited by 4,735 · see citing works
Regularizing Neural Networks by Penalizing Confident Output Distributions
Pereyra, G., Tucker, G., Chorowski, J., Kaiser, Ł., & Hinton, G. (2017) · arXiv · doi:10.48550/arxiv.1701.06548 · Cited by 655 · see citing works
Controllable Abstractive Summarization
Fan, A., Grangier, D., & Auli, M. (2018) · Proceedings of the 2nd Workshop on Neural Machine Translation and Generation · doi:10.18653/v1/w18-2706 · Cited by 281 · cited 2× · see citing works
“When generating summaries, we follow standard practice of tuning the maximum output length and disallow repeating the same trigram (Paulus et al., 2017; Fan et al., 2017).”
MethodSelf-citation
Flagged first, then by citations
Research works printed but not linked: 24
- Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. 2016. Layer normalization. arXiv, abs/1607.06450. · arXiv
- Alexei Baevski and Michael Auli. 2018. Adaptive in- put representations for neural language modeling. arXiv, abs/1809.10853. · arXiv
- Alexei Baevski, Sergey Edunov, Yinhan Liu, Luke Zettlemoyer, and Michael Auli. 2019. Cloze-driven pretraining of self-attention networks. arXiv.
- Andrew M. Dai and Quoc V. Le. 2015. Semi-supervised sequence learning. arXiv, abs/1511.01432. · arXiv
- Sebastian Gehrmann, Yuntian Deng, and Alexander M Rush. 2018. Bottom-up abstractive summarization. arXiv, abs/1808.10792. · arXiv
- Caglar Gulcehre, Orhan Firat, Kelvin Xu, Kyunghyun Cho, Loic Barrault, Huei-Chi Lin, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2015. On us- ing monolingual corpora in neural machine transla- tion. arXiv, abs/1503.03535. · arXiv
- Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Su- leyman, and Phil Blunsom. 2015. Teaching ma- chines to read and comprehend. In Proc. of NIPS.
- Hakan Inan, Khashayar Khosravi, and Richard Socher. 2016. Tying word vectors and word classifiers: A loss framework for language modeling. arXiv, abs/1611.01462. · arXiv
- Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Workshop on Text Sum- marization Branches Out.
- Ilya Loshchilov and Frank Hutter. 2016. SGDR: stochastic gradient descent with restarts. arXiv, abs/1608.03983. · arXiv
- Ramesh Nallapati, Bowen Zhou, Caglar Gulcehre, Bing Xiang, et al. 2016. Abstractive text summa- rization using sequence-to-sequence rnns and be- yond. In Proc. of CONLL.
- Myle Ott, Sergey Edunov, Alexei Baevski, Angela Fan, Sam Gross, Nathan Ng, David Grangier, and Michael Auli. 2019. fairseq: A fast, extensible toolkit for sequence modeling. In Proc. of NAACL System Demonstrations.
- Myle Ott, Sergey Edunov, David Grangier, and Michael Auli. 2018. Scaling neural machine trans- lation. In Proc. of WMT.
- Romain Paulus, Caiming Xiong, and Richard Socher. 2017. A deep reinforced model for abstractive sum- marization. arXiv, abs/1705.04304. · arXiv
- Matthew E Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. Deep contextualized word rep- resentations. In Proc. of ACL.
- Jason Phang, Thibault Fevry, and Samuel R. Bowman. 2018. Sentence encoders on stilts: Supplementary training on intermediate labeled-data tasks. arXiv, abs/1811.01088. · arXiv
- Matt Post. 2018. A call for clarity in reporting bleu scores. arXiv, abs/1804.08771. · arXiv
- Ofir Press and Lior Wolf. 2017. Using the output em- bedding to improve language models. In Proc. of EACL.
- Abigail See, Peter J Liu, and Christopher D Manning. 2017. Get to the point: Summarization with pointer- generator networks. In Proc. of ACL.
- Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016. Neural machine translation of rare words with subword units. In Proc. of ACL.
- Ilya Sutskever, James Martens, George E. Dahl, and Geoffrey E. Hinton. 2013. On the importance of ini- tialization and momentum in deep learning. In Proc. of ICML.
- Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna. 2015. Re- thinking the Inception Architecture for Computer Vision. arXiv preprint arXiv:1512.00567. · arXiv
- Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention Is All You Need. In Proc. of NIPS.
- Barret Zoph, Deniz Yuret, Jonathan May, and Kevin Knight. 2016. Transfer learning for low-resource neural machine translation. In Proc. of EMNLP. A Detailed WMT English-German Results bitext method 2012 2013 2014 2015 2016 2017 2018 Avg 160K baseline 13.2 15.7 13.5 15.7 18.6 14.8 21.4 16.1 SHARED 15.3 18.2 16.7 19.0 21.6 18.2 24.9 19.1 SHARED+bitext-BPE 15.1 17.9 16.2 18.9 22.0 18.0 25.2 19.0 SRC-ELMO 16.0 19.4 17.1 19.9 23.0 18.7 26.6 20.1 SRC-FT 15.3 18.5 16.6 18.9 20.8 17.6 24.3 18.9 TGT-ELMO 13.3 16.4 14.1 16.2 18.8 14.9 21.3 16.4 TGT-FT 14.7 17.2 15.8 18.4 21.4 16.9 24.2 18.4 SRC-ELMO+SHDEMB 17.4 20.8 18.6 21.5 24.9 20.3 29.0 21.8 320K baseline 17.2 20.4 18.1 21.2 25.0 19.6 28.9 21.5 SHARED 18.1 21.1 19.1 22.4 26.3 21.2 30.6 22.7 SHARED+bitext-BPE 17.6 20.6 19.1 22.3 26.1 20.8 29.9 22.3 src-elmo 18.8 22.3 21.1 24.0 27.5 22.2 32.5 24.1 SRC-FT 19.0 22.5 20.9 23.5 26.9 22.2 32.1 23.9 TGT-ELMO 16.7 20.7 18.2 20.9 24.1 19.4 28.0 21.1 TGT-FT 16.1 19.4 17.1 20.0 23.1 18.5 26.3 20.1 SRC-ELMO+SHDEMB 19.5 22.9 21.2 24.0 27.4 22.4 32.3 24.2 640K baseline 19.2 22.9 21.2 24.5 27.9 22.4 33.1 24.5 SHARED 19.9 23.4 22.1 25.1 28.8 23.0 34.1 25.2 SHARED+bitext-BPE 19.4 22.8 21.7 24.9 28.4 22.9 33.6 24.8 src-elmo 21.0 24.3 23.4 26.5 30.0 24.6 35.6 26.5 SRC-FT 20.5 24.0 22.9 26.1 29.1 24.4 34.9 26.0 TGT-ELMO 18.9 22.6 20.8 24.2 27.5 22.3 31.9 24.0 TGT-FT 18.2 21.8 20.6 23.7 27.0 21.8 31.4 23.5 SRC-ELMO+SHDEMB 21.2 25.1 23.9 26.7 30.2 24.7 36.2 26.9 1280K baseline 20.9 24.6 23.6 26.5 30.5 24.7 36.2 26.7 SHARED 21.1 24.6 24.6 27.6 31.0 25.2 37.3 27.3 SHARED+bitext-BPE 20.5 24.0 23.9 26.2 30.6 24.8 36.2 26.6 src-elmo 22.1 25.7 25.7 28.5 31.7 26.3 38.2 28.3 SRC-FT 21.3 25.2 25.3 28.5 31.1 26.2 37.4 27.9 TGT-ELMO 20.9 24.4 23.6 26.6 30.3 24.9 36.1 26.7 TGT-FT 20.1 23.7 22.4 25.2 29.1 23.6 34.4 25.5 SRC-ELMO+SHDEMB 22.3 26.0 26.3 28.9 32.6 26.8 38.6 28.8 2560K baseline 21.7 25.6 25.4 28.2 32.3 26.2 39.1 28.4 SHARED 22.2 25.9 25.7 28.3 32.1 26.3 38.9 28.5 SHARED+bitext-BPE 21.8 25.5 25.5 27.9 32.1 26.0 38.6 28.2 src-elmo 22.9 27.0 27.0 30.0 33.4 28.0 40.0 29.8 SRC-FT 22.2 26.4 26.3 29.5 32.4 27.3 39.3 29.1 TGT-ELMO 21.8 25.7 25.8 28.5 32.3 26.6 39.3 28.6 TGT-FT 21.5 25.3 24.5 27.0 30.2 25.2 36.8 27.2 SRC-ELMO+SHDEMB 23.1 27.2 27.1 29.7 33.7 27.9 40.0 29.8 5186K baseline 23.1 26.8 27.7 30.1 33.6 27.9 40.1 29.9 SHARED 22.6 26.6 27.7 30.5 33.4 28.2 40.2 29.9 SHARED+bitext-BPE 22.5 26.0 27.0 29.7 33.4 27.7 40.6 29.6 src-elmo 23.7 27.8 28.7 31.1 34.5 29.2 41.8 31.0 SRC-FT 23.1 27.0 27.8 30.5 33.7 28.3 40.8 30.2 TGT-ELMO 22.9 26.6 26.9 29.5 33.8 27.7 40.5 29.7 TGT-FT 22.3 26.1 26.1 28.9 32.5 26.5 38.8 28.7 SRC-ELMO+SHDEMB 23.4 28.0 28.8 31.2 34.5 28.7 41.8 30.9 Table 4: BLEU on newstest2012 to newstest2018 of WMT English-German translation in varioius simulated bitext size scenarios (cf. Figure 1).
Cited by
Sorted by citations
Lewis, M., Liu, Y., Goyal, N., Ghazvininejad, M., Mohamed, A., Levy, O., Stoyanov, V., & Zettlemoyer, L. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.703 · Cited by 8,821 · see citing works
Pre-trained models for natural language processing: A survey
Qiu, X., Sun, T., Xu, Y., Shao, Y., Dai, N., & Huang, X. (2020) · Science China Technological Sciences · doi:10.1007/s11431-020-1647-3 · Cited by 1,591 · see citing works
wav2vec: Unsupervised Pre-Training for Speech Recognition
Schneider, S., Baevski, A., Collobert, R., & Auli, M. (2019) · Interspeech 2019 · doi:10.21437/interspeech.2019-1873 · Cited by 1,290 · see citing works
Self-citationWu, J., Wu, X., Qiu, Z., Li, M., Lin, S., Zhang, Y., Zheng, Y., Yuan, C., & Yang, J. (2024) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocae079 · Cited by 41 · see citing works
- Authors
- Edunov, S., Baevski, A., & Auli, M.
- Affiliations
- Meta (United States) (US)
- ORCID
- Auli, Michael
- Venue
- Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers)
- Published
- 2019
- Pages
- 4052-4059
- OpenAlex
- W2922709902
- Cited by
- 122
- References
- 7
- Citations by year
- 2026: 1 · 2025: 5 · 2024: 13 · 2023: 22 · 2022: 11 · 2021: 29 · 2020: 27 · 2019: 14
- FWCI
- 7.75
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Keywords
- pre-trained language models, language generation, text generation, representation learning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
