Filtering for medical news items using a machine learning approach
Wanhong Zheng1, Evangelos Milios, Carolyn Watters
1Faculty of Computer Science, Dalhousie University, Halifax, NS B3H 1T8, Canada.
Proceedings. AMIA Symposium
|December 5, 2002
Summary
This study developed a machine learning approach to filter medical news. The system accurately distinguishes medical from non-medical articles, aiding targeted audience delivery.
Area of Science:
- Computational linguistics
- Medical informatics
- Machine learning
Background:
- Filtering medical news for specific audiences presents challenges in feature extraction.
- Automated classification of medical content is needed for efficient information dissemination.
Purpose of the Study:
- To develop and evaluate a machine learning approach for filtering medical news articles.
- To classify articles into non-medical, expert-level medical, and general medical categories.
Main Methods:
- Utilized supervised machine learning techniques: Decision Trees and Naïve Bayes.
- Trained classifiers on a dataset pre-classified by medical experts and human readers.
- Focused on extracting domain-appropriate feature sets for classification.
Main Results:
- Achieved an overall classification accuracy of approximately 78%.
- Demonstrated high accuracy (92%) in distinguishing non-medical from medical articles.
- Successfully classified articles into three distinct groups: non-medical, expert-focused medical, and general medical.
Conclusions:
- Machine learning, specifically Decision Trees and Naïve Bayes, is effective for medical news filtering.
- The approach shows promise for accurately segmenting medical content for targeted audiences.
- High accuracy in differentiating medical from non-medical content supports practical application.
