Zero-friction Adverse Drug Reaction Reporting from Clinical Instant Messaging Using Hybrid NLP

Feb 12, 2026·

Dongxu Wang

Equal Contribution

Zihong Lu

Equal Contribution

Wenbo Yuan

Kaiqiang Yuan

Di Yin

Ying Yao

Corresponding Author

Sunmin Jiang

Corresponding Author

· 1 min read

Source Document DOI DOI PDF

Abstract

This manuscript presents a proof-of-concept hybrid NLP pipeline for adverse drug reaction (ADR) signal detection and information extraction from clinical instant messaging data. It combines a keyword-based rule engine with a locally deployed LLM in a zero-shot setting to support low-friction pharmacovigilance reporting workflows.

Type

Working-Paper

Publication

SSRN preprint (under review at Drug Safety)

Submission Status

Available as an SSRN preprint and under review at Drug Safety.

SSRN: https://ssrn.com/abstract=6236659

DOI: https://dx.doi.org/10.2139/ssrn.6236659

Notes

Preprint citation: Wang, Dongxu and Lu, Zihong and Yuan, Wenbo and Yuan, Kaiqiang and Yin, Di and Yao, Yin and Jiang, Sumin. Zero-Friction Adverse Drug Reaction Reporting From Clinical Instant Messaging Using Hybrid NLP (February 12, 2026). Available at SSRN: https://ssrn.com/abstract=6236659.
Topic: ADR detection and reporting from clinical instant messaging using a hybrid rule + LLM pipeline.