Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: May 24, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Generating Human-Readable Labels for SNOMED CT Expressions with LLMs: A Study on Model Performance and Rater

Adel Bensahla^1,2, Jamil Zaghir^1,2, Yuanyuan Zheng^1,2

¹Division of Medical Information Sciences, Geneva University Hospitals, Geneva, Switzerland.

Studies in Health Technology and Informatics

|May 23, 2026

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

A gold-standard French-language annotated corpus of oncological entities with ICD-O normalisation.

Scientific data·2026

Same author

Clarifying the relationship between biomedical and health informatics and digital health: expert perspectives.

BMJ health & care informatics·2026

Same author

QT prolongation alerts lead to monitoring but rarely to therapeutic changes: a prospective hospital study.

Frontiers in pharmacology·2026

Same author

Can LLMs Turn French PET/CT Narrative Reports into Structured Knowledge?

Studies in health technology and informatics·2026

Same author

Explainable Framework for Ontology-Based Similarity: A Use Case on SNOMED CT.

Studies in health technology and informatics·2026

Same author

Advancing Knowledge in Evaluating the Clinical Impact of Large Language Models for Clinical Text Summarization: A Narrative Review.

Studies in health technology and informatics·2026

Large language models (LLMs) can generate human-readable labels for post-coordinated SNOMED CT expressions. While LLM labels are semantically equivalent to official terms, expert agreement on novel expressions is low, highlighting task subjectivity.

Area of Science:

Medical Informatics
Natural Language Processing
Clinical Terminology

Background:

Post-coordinated SNOMED CT expressions lack human-readable labels, limiting their utility in data analysis and machine learning.
Developing automated methods for labeling these expressions is crucial for broader adoption and application.

Purpose of the Study:

To evaluate the quality of labels for post-coordinated SNOMED CT expressions generated by a large language model (LLM).
To assess the semantic equivalence and phrasing quality of LLM-generated labels compared to official SNOMED CT terms.

Main Methods:

A large language model (LLM) generated labels for 100 post-coordinated SNOMED CT expressions.
Two clinical terminologists blindly rated the LLM-generated labels for semantic equivalence and phrasing quality.

Keywords:

Large Language Model (LLM)Representation Learning SNOMED CT post-coordination

Related Experiment Videos

Last Updated: May 24, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Main Results:

LLM-generated labels demonstrated semantic equivalence comparable to official SNOMED CT terms.
Errors in LLM labels were primarily stylistic, not conceptual.
Interrater reliability between experts was negligible when rating novel post-coordinated expressions.

Conclusions:

Automated labeling using LLMs is a viable approach for SNOMED CT expressions.
The main challenge lies in the inherent subjectivity of labeling novel expressions, not in model performance.
The goal should shift towards achieving semantic plausibility at scale rather than a single expert consensus.