Elamur Research⌕
☰
home / papers / 2019 / Structural Scaffolds for Citation Intent Classification in Scientific Publications

Structural Scaffolds for Citation Intent Classification in Scientific Publications

Summary

The paper addresses citation intent classification: determining why a scientific paper cites another work, for example to give context, to borrow a method, or to compare findings. Earlier systems depended on hand-crafted features such as cue phrases, bootstrapped patterns and external linguistic tools. The authors instead learn representations directly from the citation context. Tokens are embedded with GloVe and ELMo, encoded by a BiLSTM and pooled with dot-product attention. On top of this shared encoder sit the main intent classifier and two auxiliary "structural scaffolds". The first predicts the title of the section where a citation appears. The second judges whether a sentence ought to contain a citation. Labels for both auxiliary tasks arise naturally when papers are written, so large training sets can be built without manual annotation. All tasks are trained jointly with weighted losses. The model sets a new state of the art on ACL-ARC. The authors also release SciCite, a larger multi-domain dataset with a coarse category scheme. The results show that cheap, freely available signals about document structure can substantially improve a task with scarce labeled data.

Key points

  • On ACL-ARC, the model reaches 67.9% macro F1, an absolute gain of 13.3 points over the feature-based system of Jurgens et al. (2018), without hand-engineered features.
  • SciCite is introduced as a new dataset that is at least five times larger than earlier citation intent corpora and spans several scientific fields. It uses three intents: Background, Method and Result comparison. ACL-ARC uses six categories, including Uses, Extends, Motivation and Future work.
  • Architecture: GloVe + ELMo embeddings feed a BiLSTM with attention and per-task MLP/softmax heads. Training uses balanced mini-batches across tasks, the AdaDelta optimizer, gradient clipping at 5.0, and early stopping on dev macro F1.
  • Simply adding section titles as input features gave no improvement, whereas using them as an auxiliary scaffold task did help. Encoding a wider surrounding context lowered performance.
  • Code and data are available at github.com/allenai/scicite. The test sets contain 139 citations (ACL-ARC) and 1,861 citations (SciCite).

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Natural language processing

Cited by

90 papers of our library cite it (of the 140884 whose bibliographies we read); 7 of them have a page here, 83 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/n19-1361

@inproceedings{cohan2019structural,
  title = {{Structural Scaffolds for Citation Intent Classification in Scientific Publications}},
  author = {Cohan, Arman and Ammar, Waleed and Zuylen, Madeleine van and Cady, Field},
  booktitle = {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)},
  year = {2019},
  doi = {10.18653/v1/n19-1361},
  url = {https://doi.org/10.18653/v1/n19-1361}
}

References

33 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Amjad Abu-Jbara, Jefferson Ezra, and Dragomir Radev. 2013. Purpose and polarity of citation: Towards nlp-based bibliometrics. In NAACL-HLT. find
  2. Shashank Agarwal, Lisha Choubey, and Hong Yu. 2010. Automatically classifying the role of citations in biomedical articles. In AMIA Annual Symposium Proceedings, volume 2010, page 11. American Medical Informatics Association. find
  3. Tanzila Ahmed, Ben Johnson, Charles Oppenheim, and Catherine Peck. 2004. Highly cited old papers and the reasons why they continue to be cited. part ii., the 1953 watson and crick article on the structure of dna. Scientometrics, 61(2):147–156. find
  4. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural Computation. find
  5. David Jurgens, Srijan Kumar, Raine Hoover, Dan Mc-Farland, and Dan Jurafsky. 2018. Measuring the evolution of a scientific field through citation frames. TACL, 6:391–406. find our card
  6. Loet Leydesdorff. 1998. Theories of citation? Scientometrics. find
  7. Zhi Li and Yuh-Shan Ho. 2008. Use of citation per publication as an indicator to evaluate contingent valuation research. Scientometrics. find
  8. Terttu Luukkonen. 1992. Is scientists’ publishing behaviour rewardseeking? Scientometrics. find
  9. Iz Beltagy, Arman Cohan, and Kyle Lo. 2019. Scibert: Pretrained contextualized embeddings for scientific text. CoRR, abs/1903.10676. find our card
  10. Michael J Moravcsik and Poovanalingam Murugesan. 1975. Some results on the function and quality of citations. Social studies of science, 5(1):86–92. find
  11. Steven Bird, Robert Dale, Bonnie J. Dorr, Bryan R. Gibson, Mark Thomas Joseph, Min-Yen Kan, Dongwon Lee, Brett Powley, Dragomir R. Radev, and Yee Fan Tan. 2008. The acl anthology reference corpus: A reference dataset for bibliographic research in computational linguistics. In LREC. find our card
  12. Vinod Nair and Geoffrey E Hinton. 2010. Rectified linear units improve restricted boltzmann machines. In Proceedings of the 27th international conference on machine learning (ICML-10), pages 807–814. find
  13. Rich Caruana. 1997. Multitask learning. Machine Learning, 28:41–75. find
  14. Arman Cohan and Nazli Goharian. 2015. Scientific article summarization using citation-context and article’s discourse structure. In EMNLP. find
  15. Arman Cohan and Nazli Goharian. 2017. Contextualizing citations for scientific summarization using word embeddings and domain knowledge. In SI-GIR. find
  16. Arman Cohan, Luca Soldaini, and Nazli Goharian. 2015. Matching citation text and cited spans in biomedical literature: a search-oriented approach. In HLT-NAACL. find
  17. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. CoRR, abs/1810.04805. find
  18. Cailing Dong and Ulrich Schäfer. 2011. Ensemble-style self-training on citation classification. In IJC-NLP. find
  19. Matt Gardner, Joel Grus, Mark Neumann, Oyvind Tafjord, Pradeep Dasigi, Nelson F. Liu, Matthew E. Peters, Michael Schmitz, and Luke S. Zettlemoyer. 2018. Allennlp: A deep semantic natural language processing platform. CoRR, abs/1803.07640. find our card
  20. Mark Garzone and Robert E Mercer. 2000. Towards an automated citation classifier. In Conference of the Canadian Society for Computational Studies of Intelligence, pages 337–346. Springer. find
  21. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word representation. In EMNLP, pages 1532–1543. find
  22. Matthew E. Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke S. Zettlemoyer. 2018. Deep contextualized word representations. In NAACL-HLT. find
  23. Son Bao Pham and Achim Hoffmann. 2003. A new approach for scientific citation classification using cue phrases. In Australasian Joint Conference on Artificial Intelligence, pages 759–771. Springer. find
  24. Nils Reimers and Iryna Gurevych. 2017. Reporting score distributions makes a difference: Performance study of lstm-networks for sequence tagging. In EMNLP. find our card
  25. Anna Ritchie. 2009. Citation context analysis for information retrieval. Technical report, University of Cambridge, Computer Laboratory. find
  26. Henry Small. 2018. Characterizing highly cited method and non-method papers using citation contexts: The role of uncertainty. Journal of Informetrics, 12(2):461 – 480. find
  27. Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: a simple way to prevent neural networks from overfitting. The Journal of Machine Learning Research, 15(1):1929–1958. find
  28. Mary Elizabeth Stevens and Vincent Edward Giuliano. 1965. Statistical Association Methods for Mechanized Documentation: Symposium Proceedings, Washington, 1964, volume 269. US Government Printing Office. find
  29. Xuan Su, Animesh Prasad, Min-Yen Kan, and Kazunari Sugiyama. 2018. Neural multi-task learning for citation function and provenance. CoRR, abs/1811.07351. — [arXiv:1811.07351](https://arxiv.org/abs/1811.07351) find
  30. Swabha Swayamdipta, Sam Thomson, Kenton Lee, Luke S. Zettlemoyer, Chris Dyer, and Noah A. Smith. 2018. Syntactic scaffolds for semantic structures. In EMNLP. find
  31. Simone Teufel, Advaith Siddharthan, and Dan Tidhar. 2006. Automatic classification of citation function. In EMNLP, EMNLP ’06, pages 103–110, Stroudsburg, PA, USA. Association for Computational Linguistics. find our card
  32. Howard D White. 2004. Citation analysis and discourse analysis revisited. Applied linguistics, 25(1):89–116. find
  33. Matthew D Zeiler. 2012. Adadelta: an adaptive learning rate method. arXiv preprint arXiv:1212.5701. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/cohan-2019-structural-scaffolds-citation-intent-classification/