Elamur Research⌕
☰
home / papers / 2020 / TPLinker: Single-stage Joint Extraction of Entities and Relations Through Token Pair Linking

TPLinker: Single-stage Joint Extraction of Entities and Relations Through Token Pair Linking

Summary

The paper addresses joint extraction of entities and relational triplets from text, focusing on overlapping cases where triplets share one entity (SEO) or both entities (EPO). Earlier joint models either generate triplets step by step with a decoder or first find subjects and then their objects. Both approaches train on gold context but must rely on their own predictions at test time, so errors accumulate (exposure bias). The authors propose TPLinker, a single-stage model that treats extraction as linking pairs of tokens. Its handshaking tagging scheme uses three kinds of links: entity head to entity tail, subject head to object head, and subject tail to object tail. The two subject–object link types are tagged separately for each relation. No step depends on an earlier prediction, so training and inference behave the same way. On the NYT and WebNLG benchmarks, TPLinker reports state-of-the-art results. Its largest gains are on sentences with overlapping relations or many triplets, and it also decodes faster than the CasRel baseline. This matters for building knowledge bases from realistic, densely related text.

Key points

  • The task is split into 2N+1 sequence-labeling subtasks, where N is the number of relation types. The entity-boundary matrix is shared across relations. Each tag sequence covers the upper triangle of the token-pair matrix, with length (n²+n)/2.
  • The scheme handles nested entities and both SEO and EPO overlaps by design, for example "New York" and "New York City" sharing the object "De Blasio".
  • TPLinker outperforms CasRel by 6.3 F1 points on sentences containing 5 or more triplets.
  • TPLinker_BERT decodes about 3.6× faster than CasRel_BERT because it supports batched inference:
    • NYT: 15.2 ms vs 54.0 ms per sample.
    • WebNLG: 25.6 ms vs 76.8 ms per sample.
  • On WebNLG (171 relations, 5,019 training sentences), TPLinker roughly matches CasRel. The authors attribute this to near-saturated scores and to near-synonymous relation labels in the annotation. Code is published on GitHub.

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

70 papers of our library cite it (of the 140884 whose bibliographies we read); 1 of them have a page here, 69 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2020.coling-main.138

@inproceedings{wang2020tplinker,
  title = {{TPLinker: Single-stage Joint Extraction of Entities and Relations Through Token Pair Linking}},
  author = {Wang, Yucheng and Yu, Bowen and Zhang, Yueyang and Liu, Tingwen and Zhu, Hongsong and Sun, Limin},
  booktitle = {Proceedings of the 28th International Conference on Computational Linguistics},
  year = {2020},
  doi = {10.18653/v1/2020.coling-main.138},
  url = {https://doi.org/10.18653/v1/2020.coling-main.138}
}

References

25 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Giannis Bekoulis, Johannes Deleu, Thomas Demeester, and Chris Develder. 2018. Joint entity recognition and relation extraction as a multi-head selection problem. Expert Systems with Applications. find
  2. Yee Seng Chan and Dan Roth. 2011. Exploiting syntactico-semantic structures for relation extraction. In Proceedings of ACL. find
  3. Dai Dai, Xinyan Xiao, Yajuan Lyu, Shan Dou, Qiaoqiao She, and Haifeng Wang. 2019. Joint extraction of entities and overlapping relations using position-attentive sequence labeling. In Proceedings of AAAI. find
  4. Tsu-Jui Fu, Peng-Hsuan Li, and Wei-Yun Ma. 2019. Graphrel: Modeling text as relational graphs for joint entity and relation extraction. In Proceedings of ACL. find
  5. Claire Gardent, Anastasia Shimorina, Shashi Narayan, and Laura Perez-Beltrachini. 2017. Creating training corpora for nlg micro-planning. In Proceedings of ACL. find
  6. Pankaj Gupta, Hinrich Schtze, and Bernt Andrassy. 2016. Table filling multi-task recurrent neural network for joint entity and relation extraction. In Proceedings of COLING. find
  7. Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980. find
  8. Qi Li and Heng Ji. 2014. Incremental joint extraction of entity mentions and relations. In Proceedings of ACL. find our card
  9. Xiaoya Li, Fan Yin, Zijun Sun, Xiayu Li, Arianna Yuan, Duo Chai, Mingxin Zhou, and Jiwei Li. 2019. Entity-relation extraction as multi-turn question answering. In Proceedings of ACL. find our card
  10. Ilya Loshchilov and Frank Hutter. 2016. Sgdr: Stochastic gradient descent with warm restarts. arXiv preprint arXiv:1608.03983. find
  11. Makoto Miwa and Yutaka Sasaki. 2014. Modeling joint entity and relation extraction with table representation. In Proceedings of EMNLP, pages 1858–1869. find our card
  12. Tapas Nayak and Hwee Tou Ng. 2020. Effective modeling of encoder-decoder architecture for joint entity and relation extraction. In Proceedings of AAAI. find
  13. Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. Glove: Global vectors for word representation. In Proceedings of EMNLP. find
  14. Sebastian Riedel, Limin Yao, and Andrew McCallum. 2010. Modeling relations and their mentions without labeled text. In ECML-PKDD. find
  15. Changzhi Sun, Yeyun Gong, Yuanbin Wu, Ming Gong, Daxin Jiang, Man Lan, and Sun. 2019. Joint type inference on entities and relations via graph convolutional networks. In Proceedings of ACL. find
  16. Ryuichi Takanobu, Tianyang Zhang, Jiexi Liu, and Minlie Huang. 2019. A hierarchical framework for relation extraction with reinforcement learning. In Proceedings of AAAI. find
  17. Zhen Tan, Xiang Zhao, Wei Wang, and Weidong Xiao. 2019. Jointly extracting multiple triplets with multilayer translation constraints. In Proceedings of AAAI. find
  18. Zhepei Wei, Jianlin Su, Yue Wang, Yuan Tian, and Yi Chang. 2020. A novel cascade binary tagging framework for relational triple extraction. In Proceedings of ACL. find our card
  19. Xiaofeng Yu and Wai Lam. 2010. Jointly identifying entities and extracting relations in encyclopedia text via a graphical model approach. In Proceedings of COLING, pages 1399–1407. find
  20. Bowen Yu, Zhenyu Zhang, Xiaobo Shu, Yubin Wang, Tingwen Liu, Bin Wang, and Sujian Li. 2020. Joint extraction of entities and relations based on a novel decomposition strategy. In Proceedings of ECAI. find
  21. Dmitry Zelenko, Chinatsu Aone, and Anthony Richardella. 2003. Kernel methods for relation extraction. Journal of machine learning research. find our card
  22. Xiangrong Zeng, Daojian Zeng, Shizhu He, Kang Liu, Jun Zhao, et al. 2018. Extracting relational facts by an end-to-end neural model with copy mechanism. In Proceedings of ACL. find our card
  23. Xiangrong Zeng, Shizhu He, Daojian Zeng, Kang Liu, Shengping Liu, and Jun Zhao. 2019. Learning the extraction order of multiple relational facts in a sentence with reinforcement learning. In Proceedings of EMNLP-IJCNLP. find
  24. Wen Zhang, Yang Feng, Fandong Meng, Di You, and Qun Liu. 2019. Bridging the gap between training and inference for neural machine translation. In Proceedings of ACL. find our card
  25. 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. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2020-tplinker-single-stage-joint-extraction/