Elamur Research
☰
home / papers / 2017 / Frame-Semantic Parsing with Softmax-Margin Segmental RNNs and a Syntactic Scaffold

Frame-Semantic Parsing with Softmax-Margin Segmental RNNs and a Syntactic Scaffold

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive licence to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/1706.09528

Summary

The paper addresses frame-semantic argument identification: finding and labeling the text spans that fill the roles of FrameNet frames. Earlier parsers relied on syntactic parses to prune candidate spans and to supply features, which is costly and limits recall. The authors first build a parser with no syntax at all. It is a segmental RNN, which combines biLSTM span representations with a semi-Markov CRF, trained with a softmax-margin loss that favors recall. Even without syntax, this model beats the original SEMAFOR and matches its improved successor. They then add syntax in two ways. The first is a conventional pipeline that uses dependency or phrase-structure parser features. The second is a "syntactic scaffold": a multitask auxiliary objective that learns to detect unlabeled constituents from the Penn Treebank and shares the sentence encoder with the frame parser. Both reach state-of-the-art results on FrameNet 1.5. The scaffold needs no parser at training or test time, so it offers a cheaper way to benefit from syntax.

Key points

  • Released as open-source open-SESAME (SEmi-markov Softmax-margin ArguMEnt parser); a syntax-free frame identification component is also provided.
  • Evaluated on FrameNet 1.5 (1,020 frames). Segments are capped at 20 tokens, which drops under 1% of gold arguments. Prior work found that dependency-based pruning capped recall below 72.6%.
  • The scaffold is trained on the Penn Treebank as a binary is-this-span-a-constituent task. It needs only bracketing, and this need not come from the FrameNet data.
  • With ensembling, phrase-structure features give the best F1: 68.9 with gold frames and 70.9 with frames predicted by the FitzGerald et al. (2015) system. This ties the previous state of the art, which also uses PropBank.
  • The ensembled scaffold model scores 68.3 F1 with gold frames and 70.7 with predicted frames, outperforming dependency features. Across all models, softmax-margin consistently beat plain log loss.

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: Linguistics & languages

Topics: Semantic role labeling

Cited by

71 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 67 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1706.09528 · arXiv 1706.09528

@misc{swayamdipta2017frame,
  title = {{Frame-Semantic Parsing with Softmax-Margin Segmental RNNs and a Syntactic Scaffold}},
  author = {Swayamdipta, Swabha and Thomson, Sam and Dyer, Chris and Smith, Noah A.},
  year = {2017},
  doi = {10.48550/arxiv.1706.09528},
  eprint = {1706.09528},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1706.09528}
}

References

61 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Apoorv Agarwal, Sriramkumar Balasubramanian, Anup Kotalwar, Jiehan Zheng, and Owen Rambow. 2014. Frame semantic tree kernels for social network extraction from text. In Proc. of EACL. find
  2. Daniel Andor, Chris Alberti, David Weiss, Aliaksei Severyn, Alessandro Presta, Kuzman Ganchev, Slav Petrov, and Michael Collins. 2016. Globally normalized transition-based neural networks. ArXiv:1603.06042. http://arxiv.org/abs/1603.06042. ↗ our card
  3. Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. Neural machine translation by jointly learning to align and translate. ArXiv:1409.0473. ↗ our card
  4. Collin Baker, Michael Ellsworth, and Katrin Erk. 2007. SemEval’07 Task 19: Frame semantic structure extraction. In Proc. of SemEval. find
  5. Collin F. Baker, Charles J. Fillmore, and John B. Lowe. 1998. The Berkeley FrameNet project. In Proc. of ACL. find our card
  6. Anders Björkelund, Bernd Bohnet, Love Hafdell, and Pierre Nugues. 2010. A high-performance syntactic and semantic dependency parser. In Proc. of COL-ING. find
  7. Rich Caruana. 1997. Multitask learning. Machine Learning 28(1). find
  8. Xinchi Chen, Zhan Shi, Xipeng Qiu, and Xuanjing Huang. 2017. Adversarial multi-criteria learning for chinese word segmentation. ArXiv:1704.07556. ↗
  9. Yun-Nung Chen, William Yang Wang, and Alexander I Rudnicky. 2013. Unsupervised induction and filling of semantic slots for spoken dialogue systems using frame-semantic parsing. In Proc. of ASRU-IEEE. find
  10. Ronan Collobert, Jason Weston, Léon Bottou, Michael Karlen, Koray Kavukcuoglu, and Pavel Kuksa. 2011. Natural language processing (almost) from scratch. Journal of Machine Learning Research 12:2493–2537. find our card
  11. Bob Coyne, Alex Klapheke, Masoud Rouhizadeh, Richard Sproat, and Daniel Bauer. 2012. Annotation tools and knowledge representation for a text-to-scene system. In Proc. of COLING. find
  12. Dipanjan Das, Desai Chen, André FT Martins, Nathan Schneider, and Noah A Smith. 2014. Frame-semantic parsing. Computational linguistics 40(1):9–56. find our card
  13. Dipanjan Das, Andr F. T. Martins, and Noah A. Smith. 2012. An exact dual decomposition algorithm for shallow semantic parsing with constraints. In Proc. of *SEM. http://dl.acm.org/citation.cfm?id=2387636.2387671. ↗
  14. Dipanjan Das and Noah A Smith. 2011. Semi-supervised frame-semantic parsing for unknown predicates. In Proc. of ACL. find
  15. Chris Dyer, Adhiguna Kuncoro, Miguel Ballesteros, and Noah A Smith. 2016. Recurrent neural network grammars. ArXiv:1602.07776. ↗
  16. Charles J. Fillmore. 1976. Frame semantics and the nature of language. Annals of the New York Academy of Sciences 280(1):20–32. find
  17. Nicholas FitzGerald, Oscar Täckström, Kuzman Ganchev, and Dipanjan Das. 2015. Semantic role labeling with neural network factors. In Proc. of EMNLP. find
  18. Daniel Gildea and Daniel Jurafsky. 2002. Automatic labeling of semantic roles. Computational Linguistics 28(3):245–288. find our card
  19. Kevin Gimpel and Noah A. Smith. 2010. Softmax-margin CRFs: Training log-linear models with cost functions. In Proc. of NAACL. find
  20. Matthew R. Gormley, Margaret Mitchell, Benjamin Van Durme, and Mark Dredze. 2014. Low-resource semantic role labeling. In Proc. of ACL. http://www.aclweb.org/anthology/P14-1111. ↗
  21. Alex Graves. 2012. Supervised Sequence Labelling with Recurrent Neural Networks, volume 385 of Studies in Computational Intelligence. Springer. find
  22. Alex Graves. 2013. Generating sequences with recurrent neural networks. ArXiv:1308.0850. ↗
  23. Jan Hajič, Massimiliano Ciaramita, Richard Johansson, Daisuke Kawahara, Maria Antònia Martı́, Lluı́s Màrquez, Adam Meyers, Joakim Nivre, Sebastian Padó, Jan Štěpánek, Pavel Straňák, Mihai Surdeanu, Nianwen Xue, and Yi Zhang. 2009. The CoNLL-2009 shared task: Syntactic and semantic dependencies in multiple languages. In Proc. of CoNLL. find our card
  24. Silvana Hartmann, Ilia Kuznetsov, Teresa Martin, and Iryna Gurevych. 2017. Out-of-domain framenet semantic role labeling. In Proc. of EACL. find our card
  25. Luheng He, Kenton Lee, Mike Lewis, and Luke Zettlemoyer. 2017. Deep semantic role labeling: What works and what’s next. In Proc. of ACL. find our card
  26. Karl Moritz Hermann, Dipanjan Das, Jason Weston, and Kuzman Ganchev. 2014. Semantic frame identification with distributed word representations. In Proc. of ACL. find
  27. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural Computation 9(8):1735–1780. ↗
  28. Richard Johansson and Pierre Nugues. 2008. Dependency-based syntactic-semantic analysis with PropBank and NomBank. In Proc. of CoNLL. find
  29. Yoon Kim, Carl Denton, Luong Hoang, and Alexander M Rush. 2017. Structured attention networks. arXiv:1702.00887 . ↗
  30. Diederik P. Kingma and Jimmy Ba. 2014. ADAM: A method for stochastic optimization. ArXiV:1412.6980. http://arxiv.org/abs/1412.6980. ↗
  31. Lingpeng Kong, Chris Dyer, and Noah A. Smith. 2016. Segmental Recurrent Neural Networks. In Proc. of ICLR. http://arxiv.org/abs/1511.06018. ↗
  32. Meghana Kshirsagar, Sam Thomson, Nathan Schneider, Jaime Carbonell, Noah A Smith, and Chris Dyer. 2015. Frame-semantic role labeling with heterogeneous annotations. In Proc. of NAACL. find
  33. Mike Lewis, Luheng He, and Luke Zettlemoyer. 2015. Joint A* CCG parsing and semantic role labelling. In Proc. of EMNLP. find
  34. Xavier Lluı́s and Lluı́s Màrquez. 2008. A joint model for parsing syntactic and semantic dependencies. In Proc. of CoNLL. find
  35. Liang Lu, Lingpeng Kong, Chris Dyer, Noah A. Smith, and Steve Renals. 2016. Segmental recurrent neural networks for end-to-end speech recognition. ArXiV:1603.00223. ↗
  36. Minh-Thang Luong, Quoc V Le, Ilya Sutskever, Oriol Vinyals, and Lukasz Kaiser. 2015. Multi-task sequence to sequence learning. ArXiv:1511.06114. http://arxiv.org/abs/1511.06114. ↗ our card
  37. Behrang Mohit, Nathan Schneider, Rishav Bhowmick, Kemal Oflazer, and Noah A. Smith. 2012. Recall-oriented learning of named entities in arabic wikipedia. In Proc. of EACL. http://dl.acm.org/citation.cfm?id=2380816.2380839. ↗
  38. Vinod Nair and Geoffrey E. Hinton. 2010. Rectified linear units improve restricted Boltzmann machines. In Proc. of ICML. find
  39. Jason Naradowsky, Sebastian Riedel, and David A. Smith. 2012. Improving NLP through marginalization of hidden syntactic structure. In Proc. of EMNLP. http://dl.acm.org/citation.cfm?id=2390948.2391035. ↗
  40. Graham Neubig, Chris Dyer, Yoav Goldberg, Austin Matthews, Waleed Ammar, Antonios Anastasopoulos, Miguel Ballesteros, David Chiang, Daniel Clothiaux, Trevor Cohn, Kevin Duh, Manaal Faruqui, Cynthia Gan, Dan Garrette, Yangfeng Ji, Lingpeng Kong, Adhiguna Kuncoro, Gaurav Kumar, Chaitanya Malaviya, Paul Michel, Yusuke Oda, Matthew Richardson, Naomi Saphra, Swabha Swayamdipta, and Pengcheng Yin. 2017. DyNet: The dynamic neural network toolkit. ArXiv:1701.03980. ↗ our card
  41. Ankur P. Parikh, Oscar Täckström, Dipanjan Das, and Jakob Uszkoreit. 2016. A decomposable attention model for natural language inference. ArXiV:1606.01933. ↗ our card
  42. Hao Peng, Sam Thomson, and Noah A. Smith. 2017. Deep multitask learning for semantic dependency parsing. In Proc. of ACL. find
  43. Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. GloVe: Global vectors for word representation. In Empirical Methods in Natural Language Processing (EMNLP). http://www.aclweb.org/anthology/D14-1162. ↗
  44. Vasin Punyakanok, Dan Roth, and Wen-tau Yih. 2008. The importance of syntactic parsing and inference in semantic role labeling. Computational Linguistics 34(2):257–287. find our card
  45. Michael Roth. 2017. Improvingframe semantic parsing via dependency path embeddings. Https://goo.gl/3Exmip. find
  46. Michael Roth and Mirella Lapata. 2015. Context-aware frame-semantic role labeling. Transactions of the ACL 3:449–460. find
  47. Michael Roth and Mirella Lapata. 2016. Neural semantic role labeling with dependency path embeddings. ArXiv:1605.07515. ↗
  48. Michael Roth and Kristian Woodsend. 2014. Composition of word representations improves semantic role labelling. In Proc. of EMNLP. find
  49. Josef Ruppenhofer, Michael Ellsworth, Miriam RL Petruck, Christopher R Johnson, and Jan Scheffczyk. 2010. FrameNet II: Extended theory and practice. find
  50. Sunita Sarawagi, William W Cohen, et al. 2004. Semi-markov conditional random fields for information extraction. In Proc. of NIPS. volume 17. find
  51. Mike Schuster and Kuldip K. Paliwal. 1997. Bidirectional recurrent neural networks. IEEE Transactions on Signal Processing 45(11):2673–2681. find
  52. Dan Shen and Mirella Lapata. 2007. Using semantic roles to improve question answering. In Proc. of EMNLP-CoNLL. find
  53. Nitish Srivastava, Geoffrey E Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research 15(1). find
  54. Mihai Surdeanu, Richard Johansson, Adam Meyers, Lluı́s Màrquez, and Joakim Nivre. 2008. The CoNLL-2008 shared task on joint parsing of syntactic and semantic dependencies. In Proc. of CoNLL. find our card
  55. Swabha Swayamdipta, Miguel Ballesteros, Chris Dyer, and Noah A. Smith. 2016. Greedy, joint syntactic-semantic parsing with Stack LSTMs. In Proc. of CoNLL. find
  56. Oscar Täckström, Kuzman Ganchev, and Dipanjan Das. 2015. Efficient inference and structured learning for semantic role labeling. Transactions of the ACL 3:29–41. find
  57. Ivan Titov, James Henderson, Paola Merlo, and Gabriele Musillo. 2009. Online graph planarisation for synchronous parsing of semantic and syntactic dependencies. In Proc. of IJCAI. find
  58. Kristina Toutanova, Aria Haghighi, and Christopher D. Manning. 2008. A global joint model for semantic role labeling. Computational Linguistics 34(2):161– 191. find
  59. Jason Weston, Samy Bengio, and Nicolas Usunier. 2011. Wsabie: Scaling up to large vocabulary image annotation. In Proc. of IJ-CAI. http://dx.doi.org/10.5591/978-1-57735-516-8/IJCAI11-460. ↗
  60. David Wood, Jerome S. Bruner, and Gail Ross. 1976. The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry 17(2). find
  61. Jie Zhou and Wei Xu. 2015. End-to-end learning of semantic role labeling using recurrent neural networks. In Proc. of ACL. find