Resource Classification for Medical Questions.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Information Retrieval

Background:

• Medical question answering systems require efficient resource allocation.
• Classifying question types is crucial for directing users to relevant information.

Purpose of the Study:

• To develop and evaluate methods for classifying medical question resource types.
• To enable automatic question answering systems to select optimal information resources.

Main Methods:

• Manual annotation of resource types across four medical question corpora (>5,000 questions).
• Development of a supervised machine learning model using lexical, syntactic, semantic, and topic-based features for automatic classification.

Main Results:

• Manual annotation established baseline classifications.
• Automatic classification achieved accuracies ranging from 80.9% to 92.8% across datasets.
• Identified challenges in both manual and automatic classification processes.

Conclusions:

• The proposed approach effectively classifies medical question resource types.
• Automatic classification using machine learning shows high accuracy.
• Further research is needed to address classification complexities.