Elamur Research⌕
☰
home / papers / 2026 / CENT: Context Engineering Framework for Normalization of Clinical Trial Procedures

CENT: Context Engineering Framework for Normalization of Clinical Trial Procedures

Paper · 2026 · Sanya Taneja, Ziqing Ji, Hans Verstraete et al.

Summary

Clinical trial protocols describe procedures and lab tests in inconsistent wording. Mapping these terms to standard codes, known as clinical concept normalization (CCN), is usually done manually or with brittle string matching. Existing systems also focus mostly on diseases and drugs. The authors propose CENT, a two-stage pipeline that needs no fine-tuning. First, two domain-specific encoders, SapBERT and BioBERT-mnli, retrieve a short list of candidate target codes. Then a GPT model chooses the best candidate, or "no match", using a sectioned prompt with a task description, a scoring rubric and examples. They test zero-shot, few-shot, CoT, ToT and GoT prompting on UMLS LOINC-to-CPT laboratory mappings. Evaluation uses binary metrics and also hierarchical ones (Wu-Palmer, Resnik), which credit predictions that land close to the correct code in the ontology. CENT clearly beats LLM-only prompting and reaches F1 around 0.8. Running the cheaper GPT4o-mini with advanced prompting keeps accuracy similar at much lower cost. Tests on internal protocol data suggest the approach still works on noisy real-world text. This makes it practical for tasks such as patient–trial matching.

Key points

  • Benchmark: 593 LOINC–CPT mappings from UMLS remained after filtering out deprecated codes and non-specific allergen codes. The targets were 1,584 measurement-domain CPT (v2015) codes.
  • Retrieval: combining SapBERT and BioBERT-mnli candidates gives the best coverage (recall@10 = 0.85). Raising the number of candidates per model from 5 to 10 barely changes end-to-end F1 (0.82 vs 0.83).
  • Accuracy and cost:
    • CENT with GPT4.1 + CoT reached F1 = 0.82 at about $12.89 in API costs.
    • LLM-only few-shot prompting scored F1 = 0.08 and cost $102.5.
    • With ToT prompting, GPT4o-mini cost $2.66, compared with $37.57 for GPT4.1.
  • Binary vs hierarchical metrics: Compared with CoT, ToT did worse on F1 yet outperformed it on both the Wu-Palmer and Resnik measures, meaning its errors were semantically nearer to the correct code.
  • Internal validation:
    • On 139 protocol lab/biomarker procedures mapped to a custom vocabulary, CENT reached F1 = 0.71.
    • Clinicians reviewed 59 CPT predictions: 47.5% matched their annotations exactly, 28.8% were partial (parent or sibling) matches, and 23.7% were wrong.

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

🎓University of WashingtonSeattle · United States353 works · 200,043 citations🏢Johnson & JohnsonNew Brunswick · United States2 works · 16 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksBiochemistry, Genetics and Molecular Biology · 234 worksMolecular Biology · 231 works

Cite

Cite as Sanya Taneja, Ziqing Ji, Hans Verstraete and Ali Samadani. 2026. CENT: Context Engineering Framework for Normalization of Clinical Trial Procedures. BioNLP 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.bionlp-1.60

@inproceedings{taneja2026cent,
  title = {{CENT: Context Engineering Framework for Normalization of Clinical Trial Procedures}},
  author = {Sanya Taneja, Ziqing Ji, Hans Verstraete and Ali Samadani},
  booktitle = {BioNLP 2026},
  year = {2026},
  doi = {10.18653/v1/2026.bionlp-1.60},
  url = {https://doi.org/10.18653/v1/2026.bionlp-1.60}
}

References

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

  1. American Medical Association. 2025. Current procedural terminology (cpt®). find
  2. Jacob S. Berkowitz, Apoorva Srinivasan, Jose Miguel Acitores Cortina, Yasaman Fatapour, and Nicholas P Tatonetti. 2025. Biomedical text normalization through generative modeling. Journal of Biomedical Informatics, 167:104850. find
  3. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and Torsten Hoefler. 2024. Graph of Thoughts: Solving Elaborate Problems with Large Language Models. Proceedings of the AAAI Conference… find
  4. Florian Borchert, Ignacio Llorca, and Matthieu-P Schapranow. 2024. Improving biomedical entity linking for complex entity mentions with LLM-based text simplification. Database: The Journal of Biological Databases and Curation, 2024:baae067. find
  5. Katharina Danhauser, Yingding Wang, Christoph Klein, Uta Tacke, Larissa Mantoan, Laura Aurica Ritter, Florian Heinen, Chiara Nobile, and Moritz Tacke. 2025. Using large language models to extract information from pediatric clinical reports. PLOS Digital Health, 4(7):e0000919. find
  6. Pritam Deka, Anna Jurek-Loughrey, and Deepak P. 2022. Evidence extraction to validate medical claims in fake news detection. In Health Information Science: 11th International Conference, HIS 2022, Virtual Event, October 28–30, 2022, Proceedings, page 3–15, Berlin, Heidelberg. Springer-Verlag. find
  7. Nicholas J. Dobbins. 2024. Generalizable and Scalable Multistage Biomedical Concept Normalization Leveraging Large Language Models. arXiv preprint. ArXiv:2405.15122 [cs]. — [arXiv:2405.15122](https://arxiv.org/abs/2405.15122) find
  8. Arnaud Ferré, Louise Deléger, Robert Bossy, Pierre Zweigenbaum, and Claire Nédellec. 2020. C-Norm: a neural approach to few-shot entity normalization. BMC Bioinformatics, 21(S23):579. find
  9. Arnaud Ferré and Philippe Langlais. 2023. An analysis of entity normalization evaluation biases in specialized domains. BMC Bioinformatics, 24(1):227. find
  10. Yuhe Gao, Runxue Bao, Yuelyu Ji, Yiming Sun, Chenxi Song, Jeffrey P Ferraro, and Ye Ye. 2025. Transfer learning with clinical concept embeddings from large language. AMIA Summits on Translational Science Proceedings, 2025:167. find
  11. David Kartchner, Jennifer Deng, Shubham Lohiya, Te-jasri Kopparthi, Prasanth Bathala, Daniel Domingo-Fernández, and Cassie S. Mitchell. 2023. A Comprehensive Evaluation of Biomedical Entity Linking Models. Proceedings of the Conference on Empirical Methods in Natural Language Processing. Conference… find
  12. Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T Joshi, Hanna Moazam, and 1 others. 2023. Dspy: Compiling declarative language model calls into self-improving pipelines. arXiv preprint arXiv:2310.03714.
  13. Fangyu Liu, Ehsan Shareghi, Zaiqiao Meng, Marco Basaldella, and Nigel Collier. 2021. Self-Alignment Pretraining for Biomedical Entity Representations. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pa…
  14. Ivan Lopez, Akshay Swaminathan, Karthik Vedula, Sanjana Narayanan, Fateme Nateghi Haredasht, Stephen P. Ma, April S. Liang, Steven Tate, Manoj Maddali, Robert Joseph Gallo, Nigam H. Shah, and Jonathan H. Chen. 2025. Clinical entity augmented retrieval for clinical information extraction. npj Digital… find
  15. National Library of Medicine. 2006. Loinc to cpt mapping. https://www.nlm.nih.gov/research/ umls/mapping_projects/loinc_to_cpt_map. html. Accessed: 2024-02-03. ↗
  16. National Library of Medicine. 2025. Umls rest api documentation. https://documentation.uts.nlm.nih.gov/rest/home.html. Accessed: 2026-02-03. ↗
  17. OpenAI. 2022. text-embedding-ada-002. find
  18. OpenAI. 2024. Gpt-4o mini: advancing cost-efficient intelligence. find
  19. OpenAI. 2025b. OpenAI API Pricing. find
  20. Regenstrief Institute. 2025. Loinc: A universal standard for identifying laboratory observations. find
  21. Philip Resnik. 1995. Using information content to evaluate semantic similarity in a taxonomy. Proceedings of the 14th International Joint Conference on Artificial Intelligence.
  22. Israt Jahan, Md Tahmid Rahman Laskar, Chun Peng, and Jimmy Xiangji Huang. 2024. A comprehensive evaluation of large Language models on benchmark biomedical text processing tasks. Computers in Biology and Medicine, 171:108189. find
  23. Nuno Seco, Tony Veale, and Jer Hayes. 2004. An intrinsic information content metric for semantic similarity in wordnet. In Proceedings of the 16th European Conference on Artificial Intelligence, ECAI’04, page 1089–1090, NLD. IOS Press. find
  24. Sonish Sivarajkumar, Mark Kelley, Alyssa Samolyk-Mazzanti, Shyam Visweswaran, and Yanshan Wang. 2024. An empirical evaluation of prompting strategies for large language models in zero-shot clinical natural language processing: algorithm development and validation study. JMIR Medical Informatics, 12:… find
  25. Ali Soroush, Benjamin S. Glicksberg, Eyal Zimlichman, Yiftach Barash, Robert Freeman, Alexander W. Charney, Girish N. Nadkarni, and Eyal Klang. 2024. Large Language Models Are Poor Medical Coders — Benchmarking of Medical Code Querying. NEJM AI. Publisher: Massachusetts Medical Society. find
  26. Andy Wang, Cong Liu, Jingye Yang, and Chunhua Weng. 2024. Fine-tuning large language models for rare disease concept normalization. Journal of the American Medical Informatics Association, 31(9):2076–2083. find
  27. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-thought prompting elicits reasoning in large language models. In Proceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, R…
  28. Cliff Wong, Sam Preston, Qianchu Liu, Zelalem Gero, Jaspreet Bagga, Sheng Zhang, Shrey Jain, Theodore Zhao, Yu Gu, Yanbo Xu, Sid Kiblawi, Srinivasan Yegnasubramanian, Taxiarchis Botsis, Marvin Borja, Luis M. Ahumada, Joseph C. Murray, Guo Hui Gan, Roshanthi Weerasinghe, Kristina Young, and 6 others.… find
  29. Cliff Wong, Sheng Zhang, Yu Gu, Christine Moung, Jacob Abel, Naoto Usuyama, Roshanthi Weerasinghe, Brian Piening, Tristan Naumann, Carlo Bifulco, and Hoifung Poon. 2023. Scaling Clinical Trial Matching Using Large Language Models: A Case Study in Oncology. In Proceedings of the 8th Machine Learning… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/taneja-2026-cent-context-engineering-framework-normalization/