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

Knowledge discovery and data mining to assist natural language understanding

A Wilcox1, G Hripcsak

  • 1Department of Medical Informatics, Columbia University, New York, NY, USA.

Proceedings. AMIA Symposium
|February 3, 1999
PubMed
Summary

Automated knowledge discovery tools can interpret clinical natural language processing (NLP) outputs. However, performance depends on training data quality, with physician-classified data outperforming ICD9 codes.

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

Epigenetic Signatures in Monozygotic and Dizygotic Twins Discordant for Orofacial Clefts.

medRxiv : the preprint server for health sciences·2026
Same author

Controlled evaLuation of Angiotensin Receptor Blockers for COVID-19 respIraTorY disease (CLARITY): statistical analysis plan for a randomised controlled Bayesian adaptive sample size trial.

Trials·2022
Same author

MAO inhibitory activity of bromo-2-phenylbenzofurans: synthesis, <i>in vitro</i> study, and docking calculations.

MedChemComm·2018
Same author

Estimating summary statistics for electronic health record laboratory data for use in high-throughput phenotyping algorithms.

Journal of biomedical informatics·2018
Same author

A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies.

Molecular psychiatry·2017
Same author

The effects of host plant defoliation and fertilizer application on larval growth and oviposition behaviour in cinnabar moth.

Oecologia·2017

Area of Science:

  • Clinical informatics
  • Natural Language Processing (NLP)
  • Machine Learning

Background:

  • Clinical use of NLP systems is increasing, necessitating effective interpretation methods.
  • Manual interpretation by domain experts is time-consuming and resource-intensive.
  • Knowledge discovery and data mining offer automated alternatives for rule generation.

Purpose of the Study:

  • To evaluate the use of a decision tree generator (C5.0) for creating NLP interpretation rules.
  • To assess the performance of an NLP system using automatically generated rules on chest radiograph reports.

Main Methods:

  • Utilized C5.0 decision tree algorithm to generate a rule base for an NLP system.
  • Tested the NLP system on 200 chest radiograph reports.

Related Experiment Videos

  • Compared performance using training sets classified by physicians versus ICD9 codes.

    • Main Results:

    • Rule base trained on a small physician-classified set performed comparably to lay persons but below physicians.
    • Rule base trained on a larger, ICD9-coded set performed worse than both physicians and lay persons.
    • Current method's performance is limited by the size and accuracy of the training dataset.

    Conclusions:

    • Automated rule generation for NLP interpretation shows potential but requires significant improvement.
    • The quality and classification method of the training data are critical factors influencing performance.
    • A larger, accurately classified dataset is essential for enhancing the effectiveness of this automated approach.