Elamur Research⌕
☰
home / papers / 2026 / SCoNE: a Self-Correcting and Noise-Augmented Method for Complex Biological and Chemical Named Entity Recognition

SCoNE: a Self-Correcting and Noise-Augmented Method for Complex Biological and Chemical Named Entity Recognition

Paper · 2026 · Xingyu Zhu, Claire Nédellec, Balazs Nagy et al.

Summary

Generative models for biological and chemical NER can handle entities that tagging schemes struggle with, such as nested and discontinuous mentions. However, with few annotated examples they mislearn entity structure, and a single early decoding mistake can derail the rest of the output. SCoNE, a T5-based generative approach, addresses both problems with two components.

  • Noise Augmentation (NA) module: builds instruction-style inputs that list the target entity types and append the gold results. It then progressively replaces subsets of those results with noise tokens. This multiplies training variants and pushes the model to attend to entity boundaries and types.
  • Confidence-based Self-Correction (CS) module: at inference time, it finds output tokens with low prediction probability, masks them by rules, and regenerates them. This limits error propagation and recovers missed entities.

The method was tested on CHEMDNER (CEM and fine-grained variants) and the microbial ecology corpus Florilege, across few-shot and full-data settings. It beats the baselines by 1.80 F1 on CHEMDNER and 2.73 F1 on Florilege. This matters for building biology and chemistry knowledge bases when labeled data is scarce and large LLM fine-tuning is too costly.

Key points

  • Datasets: CHEMDNER-CEM, CHEMDNER fine-grained, and Florilege. Settings range from 10-shot to full training. Baselines include T5 with Instruction and DiffusionNER.
  • Complex entities: recall on nested and discontinuous entities improves by up to 4.42 points over the strongest baselines, with a 2.69-point average gain.
  • Ablation on Florilege: removing NA cuts F1 by up to 13.69 (10-shot). Removing CS costs 2.31 F1 under full training, but has mixed effects in very low-shot runs. Full SCoNE reaches 79.06 F1 there.
  • Comparison with DiffusionNER: it edges out SCoNE on full-data CHEMDNER. It degrades on the smaller, structurally harder Florilege.
  • Limitations: no external ontologies are used, and the confidence threshold and re-masking rules are set by hand. Gains are steadiest on MLM-pretrained models. Code is released on GitHub (SCoNE-paper).

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

🎓Université Paris-SaclayGif-sur-Yvette · France48 works · 6,336 citations🎓University of SzegedSzeged · Hungary13 works · 1,507 citations🏛️Institut National de Recherche pour l'Agriculture, l'Alimentation et l'EnvironnementParis · France2 works

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksBiochemistry, Genetics and Molecular Biology · 263 worksMolecular Biology · 260 works

🧭 Topics 1Research subjects and languages the work is about.Named entity recognition · 214 works

Cite

Cite as Xingyu Zhu, Claire Nédellec, Balazs Nagy, Laszlo Vidacs and Robert Bossy. 2026. SCoNE: a Self-Correcting and Noise-Augmented Method for Complex Biological and Chemical Named Entity Recognition. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-long.41

@inproceedings{zhu2026scone,
  title = {{SCoNE: a Self-Correcting and Noise-Augmented Method for Complex Biological and Chemical Named Entity Recognition}},
  author = {Xingyu Zhu, Claire Nédellec, Balazs Nagy, Laszlo Vidacs and Robert Bossy},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.41},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.41}
}

References

24 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Hamza Bousselham, Asmaa Mourhir, and 1 others. 2024. Fine-tuning gpt on biomedical nlp tasks: an empirical evaluation. In 2024 International Conference on Computer, Electrical & Communication Engineering (ICCECE), pages 1–6. IEEE. find
  2. Minsoo Cho, Jihwan Ha, Chihyun Park, and Sanghyun Park. 2020. Combinatorial feature embedding based on cnn and lstm for biomedical named entity recognition. Journal of biomedical informatics, 103:103381. find
  3. Cagla Colak and Lknur Karadeniz. 2023. Isiksumm at biolaysumm task 1: Bart-based summarization system enhanced with bio-entity labels. In The 22nd Workshop on Biomedical Natural Language Processing and BioNLP Shared Tasks, pages 636–640. find
  4. Leyang Cui, Yu Wu, Jian Liu, Sen Yang, and Yue Zhang. 2021. Template-based named entity recognition using bart. arXiv preprint arXiv:2106.01760.
  5. Hélène Falentin, Estelle Chaix, Bedis Dridi, Philippe Bessieres, Solange Buchin, Stéphanie-Marie Deutsch, Magalie Weber, Robert Bossy, Sandra Derozier, Bruno Perret, and 1 others. 2017. Florilege: a database gathering microbial phenotypes of food interest. In 4th International Conference on Microbia… find
  6. Vipina K Keloth, Yan Hu, Qianqian Xie, Xueqing Peng, Yan Wang, Andrew Zheng, Melih Selek, Kalpana Raja, Chih Hsuan Wei, Qiao Jin, and 1 others. 2024. Advancing entity recognition in biomedicine via instruction tuning of large language models. Bioinformatics, 40(4):btae163. find
  7. J-D Kim, Tomoko Ohta, Yuka Tateisi, and Jun’ichi Tsujii. 2003. Genia corpus—a semantically annotated corpus for bio-textmining. Bioinformatics, 19(suppl_1):i180–i182. find
  8. Martin Krallinger, Florian Leitner, Obdulia Rabal, Miguel Vazquez, Julen Oyarzabal, and Alfonso Valencia. 2015. Chemdner: The drugs and chemical names extraction challenge. Journal of cheminformatics, 7(Suppl 1):S1. find
  9. Fei Li, ZhiChao Lin, Meishan Zhang, and Donghong Ji. 2021a. A span-based model for joint overlapped and discontinuous named entity recognition. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language P… find
  10. Lizhou Li, Shuang Zheng, and Quanmin Wang. 2021b. Roberta and stacked bidirectional gru for fine-grained chinese named entity recognition. In Proceedings of the 2021 6th International Conference on Mathematics and Artificial Intelligence, pages 95–100. find
  11. Jian Liu, Lei Gao, Sujie Guo, Rui Ding, Xin Huang, Long Ye, Qinghua Meng, Asef Nazari, and Dhananjay Thiruvady. 2021. A hybrid deep-learning approach for complex biochemical named entity recognition. Knowledge-Based Systems, 221:106958. find
  12. Usman Naseem, Matloob Khushi, Vinay Reddy, Sakthivel Rajendran, Imran Razzak, and Jinman Kim. 2021. Bioalbert: A simple and effective pre-trained language model for biomedical named entity recognition. In 2021 International Joint Conference on Neural Networks (IJCNN), pages 1–7. IEEE. find
  13. Nhung TH Nguyen, Makoto Miwa, and Sophia Ananiadou. 2023. Span-based named entity recognition by generating and compressing information. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 1984–1996. find
  14. Nadeesha Perera, Matthias Dehmer, and Frank Emmert-Streib. 2020. Named entity recognition and relation detection for biomedical information extraction. Frontiers in cell and developmental biology, 8:673. find
  15. Basra Jehangir, Saravanan Radhakrishnan, and Rahul Agarwal. 2023. A survey on named entity recognition—datasets, tools, and methodologies. Natural Language Processing Journal, 3:100017. find
  16. Yongliang Shen, Kaitao Song, Xu Tan, Dongsheng Li, Weiming Lu, and Yueting Zhuang. 2023. Diffusionner: Boundary diffusion for named entity recognition. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3875–3890. find
  17. Bosheng Song, Fen Li, Yuansheng Liu, and Xiangxiang Zeng. 2021. Deep learning methods for biomedical named entity recognition: a survey and qualitative comparison. Briefings in Bioinformatics, 22(6):bbab282. find
  18. Cong Sun, Zhihao Yang, Lei Wang, Yin Zhang, Hongfei Lin, and Jian Wang. 2021. Biomedical named entity recognition using bert in the machine reading comprehension framework. Journal of Biomedical Informatics, 118:103799. find
  19. Shuhe Wang, Xiaofei Sun, Xiaoya Li, Rongbin Ouyang, Fei Wu, Tianwei Zhang, Jiwei Li, and Guoyin Wang. 2023. Gpt-ner: Named entity recognition via large language models. arXiv preprint arXiv:2304.10428.
  20. Tianchi Wang. 2024. Dvdner: Dual-view learning named entity recognition via diffusion. In International Conference on Knowledge Science, Engineering and Management, pages 144–163. Springer. find
  21. Sean Welleck, Ilia Kulikov, Stephen Roller, Emily Dinan, Kyunghyun Cho, and Jason Weston. 2019. Neural text generation with unlikelihood training. In International Conference on Learning Representations.
  22. Kangjie Wu, Liqian Xu, Xinxiang Li, Youhua Zhang, Zhenyu Yue, Yujia Gao, and Yiqiong Chen. 2024. Named entity recognition of rice genes and phenotypes based on bigru neural networks. Computational biology and chemistry, 108:107977. find
  23. Lijun Wu, Xu Tan, Di He, Fei Tian, Tao Qin, Jianhuang Lai, and Tie-Yan Liu. 2018. Beyond error propagation in neural machine translation: Characteristics of language also matter. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 3602–3611, Brussels, Bel… find
  24. Hang Yan, Tao Gui, Junqi Dai, Qipeng Guo, Zheng Zhang, and Xipeng Qiu. 2021. A unified generative framework for various ner subtasks. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhu-2026-scone-self-correcting-noise-augmented/