SpanBERT: Improving Pre-training by Representing and Predicting Spans
Open accessThe DOI matches its Crossref record; 53 of the 53 references it cites were found in a registry or an index.
Abstract
We present SpanBERT, a pre-training method that is designed to better represent and predict spans of text. Our approach extends BERT by (1) masking contiguous random spans, rather than random tokens, and (2) training the span boundary representations to predict the entire content of the masked span, without relying on the individual token representations within it. SpanBERT consistently outperforms BERT and our better-tuned baselines, with substantial gains on span selection tasks such as question answering and coreference resolution. In particular, with the same training data and model size as BERT large , our single model obtains 94.6% and 88.7% F1 on SQuAD 1.1 and 2.0 respectively. We also achieve a new state of the art on the OntoNotes coreference resolution task (79.6% F1), strong performance on the TACRED relation extraction benchmark, and even gains on GLUE. 1
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“XLM [46] performs MLM on a concatenation of parallel bilingual sentence pairs, called Translation Language Modeling (TLM).”
From Pre-trained models for natural language processing: A survey (2020) · cites this paper as [46]
Experimental — accuracy not yet measuredIn the citing paper
“Conneau and Lample [46] tried to initialize the entire encoder and decoder by a multilingual pre-trained BERT model and showed a significant improvement could be achieved on unsupervised MT and English-Romanian supervised MT.”
From Pre-trained models for natural language processing: A survey (2020) · cites this paper as [46]
In the citing paper
“In the experiment, this approach can surpass the BERT-style pretraining proposed by Conneau and Lample [46] both on unsupervised MT and English-Romanian supervised MT.”
From Pre-trained models for natural language processing: A survey (2020) · cites this paper as [46]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
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 54,325
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019) · arXiv · doi:10.48550/arxiv.1907.11692 · Cited by 16,515
Self-citationDecoupled Weight Decay Regularization
Loshchilov, I., & Hutter, F. (2017) · arXiv · doi:10.48550/arxiv.1711.05101 · Cited by 12,522
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
Self-citationRecursive Deep Models for Semantic Compositionality Over a Sentiment Treebank
Socher, R., Perelygin, A., Wu, J., Chuang, J., Manning, C. D., Ng, A., & Potts, C. (2013) · Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d13-1170 · Cited by 6,878 · cited 4×
“…Understanding Evaluation (GLUE) benchmark (Wang et al., 2019) consists of 9 sentence-level classification tasks: • Two sentence-level classification tasks including CoLA (Warstadt et al., 2018) for evaluating linguistic acceptability and SST-2 (Socher et al., 2013) for sentiment classification.”
BackgroundInfluentialXLNet: Generalized Autoregressive Pretraining for Language Understanding
Yang, Z., Dai, Z., Yang, Y., Carbonell, J., Salakhutdinov, R., & Le, Q. V. (2019) · arXiv · doi:10.48550/arxiv.1906.08237 · Cited by 6,687 · cited 4×
“While others show the benefits of adding more data (Yang et al., 2019) and increasing model size (Lample and Conneau, 2019), this work demonstrates the importance of designing good pretraining tasks and objectives, which can also have a significant impact.”
BackgroundMethodInfluentialSQuAD: 100,000+ Questions for Machine Comprehension of Text
Rajpurkar, P., Zhang, J., Lopyrev, K., & Liang, P. (2016) · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d16-1264 · Cited by 6,208
Universal Language Model Fine-tuning for Text Classification
Howard, J., & Ruder, S. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1031 · Cited by 4,100 · cited 2×
“, 2018) have had immense impact on NLP lately, particularly as methods for initializing a large model before fine-tuning it for a specific task (Howard and Ruder, 2018; Radford et al., 2018; Devlin et al., 2019).”
Method
Research works printed but not linked: 7
- Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. 2016. Layer normalization. arXiv preprint arXiv:1607.06450 . · arXiv
- Andrew M Dai and Quoc V Le. 2015. Semisupervised sequence learning. In Advances in Neural Information Processing Systems (NIPS) , pages 3079-3087.
- Adam Fisch, Alon Talmor, Robin Jia, Minjoon Seo, Eunsol Choi, and Danqi Chen. 2019. MRQA 2019 shared task: Evaluating generalization in reading comprehension. In Proceedings of 2nd Machine Reading for Reading Comprehension (MRQA) Workshop at EMNLP .
- Dan Hendrycks and Kevin Gimpel. 2016. Gaussian error linear units (gelus). arXiv preprint arXiv:1606.08415 . · arXiv
- Mandar Joshi, Eunsol Choi, Omer Levy, Daniel Weld, and Luke Zettlemoyer. 2019a. pair2vec: Compositional word-pair embeddings for crosssentence inference. In North American Association for Computational Linguistics (NAACL) , pages 3597-3608.
- Liu Yinhan, 2019, arxiv preprint arXiv:1907.11692 · arXiv
- Mitchell Stern, Noam Shazeer, and Jakob Uszkoreit. 2018. Blockwise parallel decoding for deep autoregressive models. In Advances in Neural Information Processing Systems (NIPS) .
Cited by
Items 1,001–1,012 of 1,012
Fasor: A Fast Tensor Program Optimization Framework for Efficient DNN Deployment
Huang, H., Chen, X., & Zhao, J. (2024) · International Conference on Supercomputing · doi:10.1145/3650200.3656631 · Cited by 6
State-of-the-Art Review on Recent Trends in Automatic Speech Recognition
Kandji, A. K., Ba, C., & Ndiaye, S. (2024) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-031-63999-9_11 · Cited by 6
Cao, R., Zhao, Z., Un, K.-F., Yu, W.-H., Martins, R. P., & Mak, P. (2024) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2024.3462560 · Cited by 6
“For question-answering (QA) tasks, transformer-based models have consistently outperformed their counterparts on the benchmark [1], [2], [3], [4].”
BackgroundPrompt Learning for Few-Shot Question Answering via Self-Context Data Augmentation
Qiu, J.-Q., Zhang, C.-Y., & Chen, C. L. P. (2024) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2024.3483201 · Cited by 5
“We evaluate the effectiveness of our method by comparing it with PLMs, i.e., RoBERTa [16] and SpanBERT [17], fine-tuned using the span-selection objective, as well as a QA task-specific pretrained model, i.e., Splinter [38], and the state-of-the-art model, i.e, FewshotQA [39], that aligns the FT of…”
An Analysis and Mitigation of the Reversal Curse
Lv, A., Zhang, K., Xie, S., Tu, Q., Chen, Y., Wen, J.-R., & Yan, R. (2024) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2024.emnlp-main.754 · Cited by 21
“Previous studies (Raffel et al., 2020; Joshi et al., 2020) suggest that masking a contiguous span of tokens Training data "prompt": "Known for being the renowned composer of the world's first underwater symphony, "Abyssal Melodies."”
Iterative Paraphrastic Augmentation with Discriminative Span Alignment
Culkin, R., Hu, J. E., Stengel-Eskin, E., Qin, G., & Durme, B. V. (2021) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00380 · Cited by 5
“While we have chosen to make use of spanpooled BERT representations in our alignment model, a natural direction for future work would be to use span-based representations such as SpanBERT (Joshi et al., 2020).”
MethodBenlahbib, A., Alami, A., & Alami, H. (2021) · International Workshop on Semantic Evaluation (SemEval) · doi:10.18653/v1/2021.semeval-1.116 · Cited by 5
“The first solution, which follows a supervised approach, is based on SpanBERT (Joshi et al., 2020) model that is pre-trained on span boundary objec-tive and considers masks contiguous spans.”
MethodOn the Robustness of Reading Comprehension Models to Entity Renaming
Yan, J., Yang, X., Mukherjee, S., Lin, B. Y., Jia, R., & Ren, X. (2022) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2022.naacl-main.37 · Cited by 19
Unsupervised Neural Machine Translation with Universal Grammar
Li, Z., Utiyama, M., Sumita, E., & Zhao, H. (2021) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2021.emnlp-main.261 · Cited by 7
“In span-based MLM like (Joshi et al., 2020), a span of length (cid:96) is first sampled from a geometric distribution (cid:96) ∼ Geo( p ) , and the start position of a span is sampled in the same manner as in MLM, giving final masked span set M S = { ( M i , (cid:96) i ) } .”
MethodImproving the Performance of Vietnamese–Korean Neural Machine Translation with Contextual Embedding
Vu, V.-H., Nguyen, Q.-P., Tunyan, E. V., & Ock, C.-Y. (2021) · Applied Sciences · doi:10.3390/app112311119 · Cited by 6
Samuel, D. (2023) · Proceedings of the BabyLM Challenge at the 27th Conference on Computational Natural Language Learning · doi:10.18653/v1/2023.conll-babylm.19 · Cited by 5
Automatic Analysis of Substantiation in Scientific Peer Reviews
Guo, Y., Shang, G., Rennard, V., Vazirgiannis, M., & Clavel, C. (2023) · arXiv · doi:10.48550/arxiv.2311.11967 · Cited by 5
Metadata
- Authors
- Joshi, M., Chen, D., Liu, Y., Weld, D. S., Zettlemoyer, L., & Levy, O.
- Affiliations
- University of Washington (US)
- Princeton University (US)
- Meta (United States) (US)
- Venue
- Transactions of the Association for Computational Linguistics
- Publisher
- MIT Press
- Published
- 2020-12
- Volume
- 8
- Pages
- 64-77
- ISSN
- 2307-387X (ISSN-L)
- OpenAlex
- W2962753370
- Cited by
- 1,827
- References
- 51
- Citations by year
- 2026: 84 · 2025: 169 · 2024: 214 · 2023: 373 · 2022: 342 · 2021: 393 · 2020: 224
- FWCI
- 130.82
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- Locations
- Transactions of the Association for Computational Linguistics (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University)
- Funders
- University of Washington
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Keywords
- SpanBERT, BERT pre-training, span prediction, question answering, coreference resolution, relation extraction, text 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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.