Elamur Research⌕
☰
home / papers / 2016 / Bilingually-constrained Synthetic Data for Implicit Discourse Relation Recognition

Bilingually-constrained Synthetic Data for Implicit Discourse Relation Recognition

Summary

Implicit discourse relation recognition, which classifies the relation between two text spans that have no explicit connective, is held back by scarce labeled data. A common workaround deletes connectives from explicit examples to create synthetic training data. This tends to change meaning and produces text whose word distribution differs from genuine implicit examples. Wu et al. take a different route. Chinese often leaves out connectives that English translations make explicit, so they mine a Chinese-English parallel corpus for pairs where an implicit Chinese instance matches an explicit English one. The English connective then labels both sides, producing BiSynData. A simple multi-task deep averaging network learns connective prediction on BiSynData as an auxiliary task alongside relation classification. This setup gives significant gains on both the English PDTB and the Chinese CDTB. It also outperforms auxiliary tasks built on conventional synthetic data. The work matters because it presents translation divergence as a cheap, cleaner source of training signal for a low-resource discourse task.

Key points

  • BiSynData construction: The data come from about 2.38M Chinese-English sentence pairs (FBIS plus HongKong Law). English explicit instances were found with the pdtb-parser toolkit. The result is 30,032 instances per language covering 80 connectives, with and, if, as, when and but the most frequent.
  • Model (MTN): Each argument is represented by averaged, fixed word2vec embeddings trained on Gigaword. The main and auxiliary tasks share their feature layers and use separate softmax classifiers. Training alternates between the two tasks, one per epoch.
  • PDTB results (4-way, macro F1): Score rose from 38.35 to 42.52. Comparison gained the most, at about 6.4 F1. MTN_bi beats Rutherford and Xue (2015) (40.50) and Liu et al. (2016) with any single auxiliary task. It falls short of Liu et al.'s combined conn+exp+rst setup (44.98), which relies on additional labeled corpora.
  • CDTB results (3-way, macro F1): Score rose from 55.44 to 58.28. MTN_bi beats the SynData-style auxiliary tasks exp (56.42) and conn (56.86).
  • Connective filtering: Keeping only unambiguous connectives did not help, possibly because the remaining data became more unbalanced. Leaving out and helped slightly for Chinese.

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: Discourse analysis

Cited by

12 papers of our library cite it (of the 140673 whose bibliographies we read); 7 of them have a page here, 5 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/d16-1253

@inproceedings{wu2016bilingually,
  title = {{Bilingually-constrained Synthetic Data for Implicit Discourse Relation Recognition}},
  author = {Wu, Changxing and Shi, Xiaodong and Chen, Yidong and Huang, Yanzhou and Su, Jinsong},
  booktitle = {Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing},
  year = {2016},
  doi = {10.18653/v1/d16-1253},
  url = {https://doi.org/10.18653/v1/d16-1253}
}

References

26 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Or Biran and Kathleen McKeown. 2013. Aggregated Word Pair Features for Implicit Discourse Relation Disambiguation. In Proceedings of ACL (Volume 2: Short Papers), pages 69–73, Sofia, Bulgaria. Chloé Braud and Pascal Denis. 2014. Combining Natural and Artificial Examples to Improve Implicit Discours… find our card
  2. John Duchi, Elad Hazan, and Yoram Singer. 2011. Adaptive Subgradient Methods for Online Learning and Stochastic Optimization. The Journal of Machine Learning Research, 12:2121–2159. Mohit Iyyer, Varun Manjunatha, Jordan Boyd-Graber, and Hal Daumé III. 2015. Deep Unordered Composition Rivals Syntact… find
  3. Yangfeng Ji and Jacob Eisenstein. 2015. One Vector is Not Enough: Entity-Augmented Distributed Semantics for Discourse Relations. In Transactions of the Association for Computational Linguistics, volume 3, pages 329–344. find our card
  4. Yangfeng Ji, Gongbo Zhang, and Jacob Eisenstein. 2015. Closing the Gap: Domain Adaptation from Explicit to Implicit Discourse Relations. In Proceedings of EMNLP, pages 2219–2224, Lisbon, Portugal. find our card
  5. Man Lan, Yu Xu, and Zhengyu Niu. 2013. Leveraging Synthetic Discourse Data via Multi-task Learning for Implicit Discourse Relation Recognition. In Proceedings of ACL, pages 476–485, Sofia, Bulgaria. find our card
  6. Junyi Jessy Li, Marine Carpuat, and Ani Nenkova. 2014a. Cross-lingual Discourse Relation Analysis: A Corpus Study and a Semi-supervised Classification System. In Proceedings of COLING : Technical Papers, pages 577–587, Dublin, Ireland. find our card
  7. Yancui Li, Wenhe Feng, Jing Sun, Fang Kong, and Guodong Zhou. 2014b. Building Chinese Discourse Corpus with Connective-driven Dependency Tree Structure. In Proceedings of EMNLP, pages 2105–2114, Doha, Qatar. find
  8. Ziheng Lin, Min-Yen Kan, and Hwee Tou Ng. 2009. Recognizing Implicit Discourse Relations in the Penn Discourse Treebank. In Proceedings of EMNLP, pages 343–351, PA, USA. find our card
  9. Ziheng Lin, Hwee Tou Ng, and Min-Yen Kan. 2014. A PDTB-styled End-to-end Discourse Parser. Natural Language Engineering, 20(02):151–184. find
  10. Xiaodong Liu, Jianfeng Gao, Xiaodong He, Li Deng, Kevin Duh, and Ye-Yi Wang. 2015. Representation Learning Using Multi-task Deep Neural Networks for Semantic Classification and Information Retrieval. In Proceedings of NAACL, pages 912–921, Denver, Colorado. find
  11. Yang Liu, Sujian Li, Xiaodong Zhang, and Zhifang Sui. 2016. Implicit Discourse Relation Classification via Multi-Task Neural Networks. In Proceedings of AAAI, pages 2750–2756, Arizona, USA. find
  12. Annie Louis, Aravind Joshi, Rashmi Prasad, and Ani Nenkova. 2010. Using Entity Features to Classify Implicit Discourse Relations. In Proceedings of SIG-DIAL, pages 59–62, PA, USA. find our card
  13. Minh-Thang Luong, Quoc V. Le, Ilya Sutskever, Oriol Vinyals, and Lukasz Kaiser. 2016. Multi-task Sequence to Sequence Learning. In Proceedings of ICLR, pages 1–10, San Juan, Puerto Rico. find our card
  14. Christopher D. Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven J. Bethard, and David McClosky. 2014. The Stanford CoreNLP Natural Language Processing Toolkit. In Proceedings of ACL (System Demonstrations), pages 55–60, Maryland, USA. find our card
  15. Daniel Marcu and Abdessamad Echihabi. 2002. An Unsupervised Approach to Recognizing Discourse Relations. In Proceedings of ACL, pages 368–375, PA, USA. find our card
  16. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient Estimation of Word Representations in Vector Space. CoRR, abs/1301.3781. find
  17. Emily Pitler, Annie Louis, and Ani Nenkova. 2009. Automatic Sense Prediction for Implicit Discourse Relations in Text. In Proceedings of ACL-IJCNLP, pages 683–691, PA, USA. find our card
  18. Rashmi Prasad, Nikhil Dinesh, Alan Lee, Eleni Milt-sakaki, Livio Robaldo, Aravind Joshi, and Bonnie Webber. 2008. The Penn Discourse TreeBank 2.0. In Proceedings of LREC, volume 24, pages 2961–2968, Marrakech, Morocco. find our card
  19. Attapol T. Rutherford and Nianwen Xue. 2014. Discovering Implicit Discourse Relations Through Brown Cluster Pair Representation and Coreference Patterns. In Proceedings of EACL, pages 645–654, Gothenburg, Sweden. find our card
  20. Attapol Rutherford and Nianwen Xue. 2015. Improving the Inference of Implicit Discourse Relations via Classifying Explicit Discourse Connectives. In Proceedings of NAACL, pages 799–808, Denver, Colorado. find our card
  21. Radu Soricut and Daniel Marcu. 2003. Sentence Level Discourse Parsing Using Syntactic and Lexical Information. In Proceedings of NAACL, pages 149–156, PA, USA. find our card
  22. Caroline Sporleder and Alex Lascarides. 2008. Using Automatically Labelled Examples to Classify Rhetorical Relations: An Assessment. Natural Language Engineering, 14(3):369–416. find
  23. Mann William and Sandra Thompson. 1988. Rhetorical structure theory: Towards a Functional Theory of Text Organization. Text, 8(3):243–281. find
  24. Biao Zhang, Jinsong Su, Deyi Xiong, Yaojie Lu, Hong Duan, and Junfeng Yao. 2015. Shallow Convolutional Neural Network for Implicit Discourse Relation Recognition. In Proceedings of EMNLP, pages 2230– 2235, Lisbon, Portugal. find our card
  25. Yuping Zhou and Nianwen Xue. 2012. PDTB-style discourse annotation of Chinese text. In Proceedings of ACL, pages 69–77, Jeju Island, Korea. find
  26. Zhi-Min Zhou, Yu Xu, Zheng-Yu Niu, Man Lan, Jian Su, and Chew Lim Tan. 2010. Predicting Discourse Connectives for Implicit Discourse Relation Recognition. In Proceedings of COLING, pages 1507–1514, PA, USA. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wu-2016-bilingually-constrained-synthetic-data-implicit/