Elamur Research⌕
☰
home / papers / 2019 / GraphRel: Modeling Text as Relational Graphs for Joint Entity and Relation Extraction

GraphRel: Modeling Text as Relational Graphs for Joint Entity and Relation Extraction

Paper · 2019 · Tsu-Jui Fu, Peng-Hsuan Li and Wei-Yun Ma

Summary

GraphRel targets joint extraction of entities and relation triplets from raw sentences. It addresses three problems that earlier systems did not solve together: recognizing entities and relations in a single end-to-end model, handling triplets that share entity mentions (overlapping relations), and modeling how relations influence each other.

The model first encodes words with GloVe and POS embeddings fed into a Bi-LSTM. A bidirectional GCN then runs over a dependency parse, so each word captures both sequential and syntactic-neighborhood context. In this first phase, the model tags entities and scores every directed word pair for every relation type. Because predictions are made per word pair, shared mentions are no longer a barrier. In the second phase, these scores serve as edge weights in a complete graph built for each relation. A relation-weighted GCN passes information across this graph before the final classification.

On NYT and WebNLG, GraphRel raises recall substantially without sacrificing precision. It sets a new state-of-the-art, with F1 gains of 3.2% and 5.8% over earlier work.

Key points

  • Datasets and baselines: Evaluated on NYT and WebNLG against NovelTagging and the seq2seq OneDecoder/MultiDecoder models. GraphRel improves on every overlap category (normal, EntityPairOverlap, SingleEntityOverlap).
  • Gains by triplet type and count: On WebNLG, the first-phase model alone surpasses MultiDecoder by 3.5% (normal), 2.9% (EPO) and 3.4% (SEO). For sentences with many triplets, the two-phase model leads MultiDecoder by 11.1% on NYT.
  • Best configuration: Two first-phase GCN layers plus one relation-weighted layer gives relation F1 of 61.9% on NYT and 42.9% on WebNLG. Adding more layers or a third phase brings no further improvement.
  • Entity recognition: The second phase also improves entity tagging, with F1 rising from 88.8% to 89.2% on NYT and from 89.1% to 91.9% on WebNLG.
  • Threshold-based inference: This decoding method outperforms head and average decoding. Lowering θ from 0.8 to 0 increases recall by 1.8% on NYT and 1.4% on WebNLG, at a cost of only about 0.6% precision.

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

 Institutions

🔬Academia SinicaTaipei · Taiwan5 works · 737 citations

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 1Research subjects and languages the work is about.Information extraction · 109 works

Cite

Cite as Tsu-Jui Fu, Peng-Hsuan Li and Wei-Yun Ma. 2019. GraphRel: Modeling Text as Relational Graphs for Joint Entity and Relation Extraction. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/p19-1136

@inproceedings{fu2019graphrel,
  title = {{GraphRel: Modeling Text as Relational Graphs for Joint Entity and Relation Extraction}},
  author = {Tsu-Jui Fu, Peng-Hsuan Li and Wei-Yun Ma},
  booktitle = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics},
  year = {2019},
  doi = {10.18653/v1/p19-1136},
  url = {https://doi.org/10.18653/v1/p19-1136}
}

References

29 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. A. Cetoli, S. Bragaglia, A.D. O’Harney, and M. Sloan. 2016. Graph convolutional networks for named entity recognition. In Proceedings of TLT. find
  2. Yee-Seng Chan and Dan Roth. 2011. Exploiting syntactico-semantic structures for relation extraction. In Proceedings of ACL.
  3. Claire Gardent, Anastasia Shimorina, Shashi Narayan, and Laura Perez-Beltrachini. 2017. Creating training corpora for nlg micro-planners. In Proceedings of ACL.
  4. Sepp Hochreiter and Jrgen Schmidhuber. 1997. Long short-term memory. Neural Computation. find
  5. Matthew Honnibal and Mark Johnson. 2015. An improved non-monotonic transition system for dependency parsing. In Proceedings of EMNLP.
  6. Diederik P. Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. In Proceedings of ICLR. find
  7. Thomas Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. In Proceedings of ICLR.
  8. Jiwei Li, Minh-Thang Luong, Dan Jurafsky, and Eu-dard Hovy. 2015. When are tree structures necessary for deep learning of representations? In Proceedings of EMNLP.
  9. Yi Luan, Dave Wadden, Luheng He, Amy Shah, Mari Ostendorf, and Hannaneh Hajishirzi. 2019. A general framework for information extraction using dynamic span graphs. In Proceedings of NAACL.
  10. Diego Marcheggiani and Ivan Titov. 2017. Encoding sentences with graph convolutional networks for semantic role labeling. In Proceedings of EMNLP.
  11. Makoto Miwa and Mohit Bansal. 2016. End-to-end relation extraction using lstms on sequences and tree structures. In Proceedings of ACL.
  12. Makoto Miwa and Yutaka Sasaki. 2014. Modeling joint entity and relation extraction with table representation. In Proceedings of EMNLP.
  13. Nanyun Peng, Hoifung Poon, Chris Quirk, Kristina Toutanova, and Wen tau Yih. 2017. Cross-sentence n-ary relation extraction with graph lstms. In Proceedings of ACL.
  14. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word representation. In Proceedings of EMNLP.
  15. Yujie Qian, Enrico Santus, Zhijing Jin, Jiang Guo, and Regina Barzilay. 2019. Graphie: A graph-based framework for information extraction. In Proceedings of NAACL. find
  16. Xiang Ren, Zeqiu Wu, Wenqi He, Meng Qu, Clare R. Voss, Heng Ji, Tarek F. Abdelzaher, and Jiawei Han. 2017. Cotype: Joint extraction of typed entities and relations with knowledge bases. In Proceedings of WWW. find
  17. Sebastian Riedel, Limin Yao, and Andrew McCallum. 2010. Modeling relations and their mentions without labeled text. In Proceedings of ECML-PKDD. find
  18. Cicero Nogueira dos Santos, Bing Xiang, and Bowen Zhou. 2015. Classifying relations by ranking with convolutional neural networks. In Proceedings of ACL.
  19. Richard Socher, Brody Huval, Christopher D. Manning, and Andrew Y. Ng. 2012. Semantic compositionality through recursive matrix-vector spaces. In Proceedings of EMNLP-CoNLL.
  20. Xu Yan, Lili Mou, Ge Li, Yunchuan Chen, Hao Peng, and Zhi Jin. 2015. Classifying relations via long short term memory networks along shortest dependency paths. In Proceedings of EMNLP.
  21. Qi Li and Heng Ji. 2014. Incremental joint extraction of entity mentions and relations. In Proceedings of ACL.
  22. Xiaofeng Yu and Wai Lam. 2010. Jointly identifying entities and extracting relations in encyclopedia text via a graphical model approach. In Proceedings of COLING. find
  23. Bang Liu, Ting Zhang, Di Niu, Jinghong Lin, Kunfeng Lai, and Yu Xu. 2018. Matching long text documents via graph convolutional networks. arXiv preprint arXiv:1802.07459. — [arXiv:1802.07459](https://arxiv.org/abs/1802.07459) find
  24. Dmitry Zelenko, Chinatsu Aone, and Anthony Richardella. 2003. Kernel methods for relation extraction. Journal of Machine Learning Research (JMLR).
  25. Daojian Zeng, Kang Liu, Siwei Lai, Guangyou Zhou, and Jun Zhao. 2014. Relation classification via convolutional deep neural network. In Proceedings of COLING.
  26. Xiangrong Zeng, Daojian Zeng, Shizhu He, Kang Liu, and Jun Zhao. 2018. Extracting relational facts by an end-to-end neural model with copy mechanism. In Proceedings of ACL.
  27. Yuhao Zhang, Peng Qi, and Christopher D. Manning. 2018. Graph convolution over pruned dependency trees improves relation extraction. In Proceedings of EMNLP.
  28. Suncong Zheng, Feng Wang, Hongyun Bao, Yuexing Hao, Peng Zhou, and Bo Xu. 2017. Joint extraction of entities and relations based on a novel tagging scheme. In Proceedings of ACL.
  29. GuoDong Zhou, Jian Su, Jie Zhang, and Min Zhang. 2005. Exploring various knowledge in relation extraction. In Proceedings of ACL.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/fu-2019-graphrel-modeling-text-as-relational/