Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Combining an expert-based medical entity recognizer to a machine-learning system: methods and a case study.

Pierre Zweigenbaum1, Thomas Lavergne, Natalia Grabar

  • 1LIMSI-CNRS, Orsay, France.

Biomedical Informatics Insights
|September 21, 2013
PubMed
Summary
This summary is machine-generated.

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

Improving Explainability in Clinical Mortality Prediction Using Stacking Classifiers over Annotated Clinical Notes.

Studies in health technology and informatics·2026
Same author

<sup>1</sup>H-NMR Analysis of Wine Metabolites: Method Development and Validation.

Molecules (Basel, Switzerland)·2026
Same author

Search-optimized quantization in biomedical ontology alignment.

Frontiers in artificial intelligence·2025
Same author

Cross- & multi-lingual medication detection: a transformer-based analysis.

BMC medical informatics and decision making·2025
Same author

Challenges in Multilingual Adverse Drug Reaction Detection on Social Media: Insights from Case Studies.

Studies in health technology and informatics·2025
Same author

Efficient Semantic Similarity Computing with Optimized BERT Models.

Studies in health technology and informatics·2025

Combining expert-based and data-driven medical entity recognition systems risks overfitting. Integrating expert knowledge into a data-driven model improved performance, but further research is needed for generalization.

Area of Science:

  • Natural Language Processing
  • Computational Linguistics
  • Bioinformatics

Background:

  • Medical entity recognition is crucial for information extraction in healthcare.
  • Current methods predominantly use supervised machine learning (data-driven).
  • Expert-based systems incorporate linguistic and domain knowledge, often complementing data-driven approaches.

Purpose of the Study:

  • To investigate the combination of an expert-based system (Ogmios) and a data-driven system (Caramba) for medical entity recognition.
  • To analyze the impact of overfitting in expert-based systems on combined model performance.
  • To evaluate a novel integration method using a Conditional Random Field (CRF) classifier.

Main Methods:

  • A case study combining Ogmios (expert-based) with Caramba (data-driven CRF classifier).
Keywords:
hybrid methodsinformation extractionmachine learningmedical recordsnatural language processingoverfitting

Related Experiment Videos

  • Feature-level analysis to understand system interactions and identify overfitting.
  • Evaluating performance using precision, recall, and F-measure metrics.

    • Main Results:

    • Direct combination of Ogmios and Caramba did not yield performance improvements due to expert system overfitting.
    • Wrapping the expert-based system as features for the CRF classifier significantly boosted its F-measure from 0.603 to 0.710.
    • The integrated expert features matched the performance of the purely data-driven system.

    Conclusions:

    • Expert-based systems can be vulnerable to overfitting, hindering performance gains when naively combined with data-driven methods.
    • Integrating expert knowledge as features within a CRF classifier is an effective strategy to enhance medical entity recognition.
    • Further investigation is required to assess the generalizability of this integration technique across different datasets and tasks.